Digital staining involves using methods such as Machine Learning (ML) to replace chemical staining in pathology. Staining adds contrast that makes cell details more visible under the microscope. However, chemical methods are slow, use toxic reagents, and require skilled personnel. In contrast, digital staining can generate images faster, reduce the need for reagents and specialized equipment, and minimize plastic and chemical waste, making the workflow more sustainable. This paper systematically reviews papers published on ML-based digital stain processing. We propose a new taxonomy that divides existing studies into five groups: stain normalization, stain augmentation, virtual staining, stain transformation, and hybrid approaches. In addition, we observed several trends from the reviewed papers. Finally, we outline open research directions.

Despite recent significant advancements in Handwritten Document Recognition (HDR), the efficient and accurate recognition of text against complex backgrounds, diverse handwriting styles, and varying document layouts remains a practical challenge. Moreover, this issue is seldom addressed in academic research, particularly in scenarios with minimal annotated data available. In this paper, we introduce the DocTTT framework to address these challenges. The key innovation of our approach is that it uses test-time training to adapt the model to each specific input during testing. We propose a novel Meta-Auxiliary learning approach that combines Meta-learning and self-supervised Masked Autoencoder (MAE). During testing, we adapt the visual representation parameters using a self-supervised MAE loss. During training, we learn the model parameters using a meta-learning framework, so that the model parameters are learned to adapt to a new input effectively. Experimental results show that our proposed method significantly outperforms existing state-of-the-art approaches on benchmark datasets.

In the field of historical manuscript research, scholars frequently encounter novel symbols in ancient texts, investing considerable effort in their identification and documentation. Although existing object detection methods achieve impressive performance on known categories, they struggle to recognize novel symbols without retraining. To address this limitation, we propose a Visually Guided Text Spotting (VGTS) approach that accurately spots novel characters using just one annotated support sample. The core of VGTS is a spatial alignment module consisting of a Dual Spatial Attention (DSA) block and a Geometric Matching (GM) block. The DSA block aims to identify, focus on, and learn discriminative spatial regions in the support and query images, mimicking the human visual spotting process. It first refines the support image by analyzing inter-channel relationships to identify critical areas, and then refines the query image by focusing on informative key points. The GM block, on the other hand, establishes the spatial correspondence between the two images, enabling accurate localization of the target character in the query image. To tackle the example imbalance problem in low-resource spotting tasks, we develop a novel torus loss function that enhances the discriminative power of the embedding space for distance metric learning . To further validate our approach, we introduce a new dataset featuring ancient Dongba hieroglyphics (DBH) associated with the Naxi minority of China. Extensive experiments on the DBH dataset and other public datasets, including Egyptian Hieroglyph (EGY), Historical Arabic Documents (HAD), Tripitaka Koreana in Han (TKH), and Notary Charters (NC), show that VGTS consistently surpasses state-of-the-art methods. The proposed framework exhibits great potential for application in historical manuscript text spotting, enabling scholars to efficiently identify and document novel symbols with minimal annotation effort.

Humans often require only a few visual archetypes to spot novel objects. Based on this observation, we present a strategy rooted in ``spotting the unseen" by establishing dense correspondences between potential query image regions and a visual archetype, and we propose the Consensus Network (CoNet). Our method leverages relational patterns intra and inter images via Auto-Correlation Representation (ACR) and Mutual-Correlation Representation (MCR). Within each image, the ACR module is capable of encoding both local self-similarity and global context simultaneously. Between the query and support images, the MCR module computes the cross-correlation across two image representations and introduces a reciprocal consistency constraint, which can incorporate to exclude outliers and enhance model robustness. To overcome the challenges of low-resource training data, particularly in one-shot learning scenarios, we incorporate an adaptive margin strategy to better handle diverse instances. The experimental results indicate the effectiveness of the proposed method across diverse domains such as object detection in natural scenes, and text spotting in both historical manuscripts and natural scenes, which demonstrates its sparkling generalization ability. Our code is available at: https://github.com/infinite-hwb/conet.

Many factors can influence the process of detecting and classifying stores based on their visual appearance. Previous studies built models that considered the whole storefront however, the detection and classification results were negatively impacted because of the lack of consistency in storefront design. This research focuses on store signboards as they are much more consistent. A complete framework is provided in which it enables existing real-time object detectors to integrate with another model connected to an OCR and then classify shops using NLP techniques. The models were trained and evaluated utilizing the ShoS dataset which was collected from Google Street Views for different research purposes. A total of 10k storefront signboards were captured and fully annotated. The outcomes of different baseline methodology and applications on the ShoS dataset are provided to measure the performance of our work.

This paper considers different deep learning architectures to learn patterns that are objects lying on the Riemannian and Grassmann manifolds. Among them, we considered cascades of classifier ensembles (CCEs), convolutional neural networks (CNNs), and deep neural forests (DNFs). All aforementioned architectures have linearized and nonlinearized versions. Patterns that are objects of Riemannian manifolds are classifier prediction pairwise matrices (CPPMs) while objects of the Grassmann manifolds are obtained using decision profiles (DPs). We also compared our architectures with CCEs that operate in the Euclidean geometry. As seen from the experimental results deep learning architectures based on CNNs provided the best results.

In vision-driven Intelligent Transportation Systems (ITS) where cameras play a vital role, accurate detection and re-identification of vehicles are fundamental demands. Hence, recent approaches have employed a wide range of algorithms to provide the best possible accuracy. These methods commonly generate a vehicle detection model based on its visual appearance features such as license plate, headlights, or some other distinguishable specifications. Among different object detection approaches, Deep Neural Networks (DNNs) have the advantage of magnificent detection accuracy in case a huge amount of training data is provided. In this paper, a robust approach for license plate detection (LPD) based on YOLO v.3 is proposed which takes advantage of high detection accuracy and real-time performance. The mentioned approach can detect the license plate location of vehicles as a general representation of vehicle presence in images. To train the model, a dataset of vehicle images with Iranian license plates has been generated by the authors and augmented to provide a wider range of data for test and train purposes. It should be mentioned that the proposed method can detect the license plate area as an indicator of vehicle presence with no Optical Character Recognition (OCR) algorithm to distinguish characters inside the license plate. Experimental results have shown the high performance of the system with a precision 0.979 and recall 0.972.

This paper introduces a new topic and research of geometric classifier ensemble learning using two types of objects: classifier prediction pairwise matrix (CPPM) and decision profiles (DPs). Learning from CPPM requires using Riemannian manifolds (R-manifolds) of symmetric positive definite (SPD) matrices. DPs can be used to build a Grassmann manifold (G-manifold). Experimental results show that classifier ensembles and their cascades built using R-manifolds are less dependent on some properties of individual classifiers (e.g. depth of decision trees in random forests (RFs) or extra trees (ETs)) in comparison to G-manifolds and Euclidean geometry. More independent individual classifiers allow obtaining R-manifolds with better properties for classification. Generally, the accuracy of classification in nonlinear geometry is higher than in Euclidean one. For multi-class problems, G-manifolds perform similarly to stacking-based classifiers built on R-manifolds of SPD matrices in terms of classification accuracy.

Automatic analysis of facial beauty has become an emerging computer vision problem in recent years. Facial beauty prediction (FBP) aims at developing a human-like model that automatically makes facial attractiveness predictions. In this study, we present and evaluate a face attractiveness prediction approach using facial parts as well as a multi-task learning scheme. First, a deep convolutional neural network (CNN) pre-trained on massive face datasets is utilized for face attractiveness prediction, which is capable of automatic learning of high-level face representations. Next, we extend our deep model to other facial attribute recognition tasks. Hence, a multi-task learning scheme is leveraged by our deep model to learn optimal shared features for three correlated tasks (i.e. facial beauty assessment, gender recognition as well as ethnicity identification). To further enhance the attractiveness computation accuracy, specific regions of face images (i.e. left eye, nose and mouth) as well as the whole face are fed into multi-stream CNNs (i.e. three two-stream networks). Each two-stream network adopts a facial part as well as the full face as input. Extensive experiments are conducted on the SCUT-FBP5500 benchmark dataset, where our approach indicates significant improvement in accuracy over the other state-of-the-art methods.

Identity-preserved face beauty transformation aims to change the beauty scale of a face image while preserving the identity of the original face. In our framework of conditional Generative Adversarial Networks (cGANs), the synthesized face produced by the generator would have the same beauty scale indicated by the input condition. Unlike the discrete class labels used in most cGANs, the condition of target beauty scale in our framework is given by a continuous real-valued beauty score in the range [1 to 5], which makes the work challenging. To tackle the problem, we have implemented a triple structure, in which the conditional discriminator is divided into a normal discriminator and a separate face beauty predictor. We have also developed another new structure called Conditioned Instance Normalization to replace the original concatenation used in cGANs, which makes the combination of the input image and condition more effective. Furthermore, Self-Consistency Loss is introduced as a new parameter to improve the stability of training and quality of the generated image. In the end, the objectives of beauty transformation and identity preservation are evaluated by the pretrained face beauty predictor and state-of-the-art face recognition network. The result is encouraging and it also shows that certain facial features could be synthesized by the generator according to the target beauty scale, while preserving the original identity.

This paper considers three different classifier stacking algorithms: simple stacking, cascades of classifier ensembles and nonlinear version of classifier stacking based on classifier interactions. Classifier interactions can be expressed using classifier prediction pairwise matrix (CPPM). As a meta-learner for the last algorithm Convolutional Neural Networks (CNNs) and two other classifier stacking algorithms (simple classifier stacking and cascades of classifier ensembles) have been applied. This allows applying classical stacking and cascade-based recursive stacking in the Euclidean and the Riemannian geometries. The cascades of random forests (RFs) and extra trees (ETs) are considered as a forest-based alternative to deep neural networks [1]. Our goal is to compare accuracies of the cascades of RFs and CNN-based stacking or deep multi-layer perceptrons (MLPs) for different classifications problems. We use gesture phase dataset from UCI repository [2] to compare and analyze cascades of RFs and extra trees (ETs) in both geometries and CNN-based version of classifier stacking. This data set was selected because generally motion is considered as a nonlinear process (patterns do no lie in Euclidean vector space) in computer vision applications. Thus we can assess how good are forest-based deep learning and the Riemannian manifolds (R-manifolds) when applied to nonlinear processes. Some more datasets from UCI repository were used to compare the aforementioned algorithms to some other well-known classifiers and their stacking-based versions in both geometries. Experimental results show that classifier stacking algorithms in Riemannian geometry (R-geometry) are less dependent on some properties of individual classifiers (e.g. depth of decision trees in RFs or ETs) in comparison to Euclidean geometry. More independent individual classifiers allow to obtain R-manifolds with better properties for classification. Generally, accuracy of classification using classifier stacking in R-geometry is higher than in Euclidean one.

The accuracies for many pattern recognition tasks have increased rapidly year by year, achieving or even outperforming human performance. From the perspective of accuracy, pattern recognition seems to be a nearly solved problem. However, once launched in real applications, the high-accuracy pattern recognition systems may become unstable and unreliable due to the lack of robustness in open and changing environments. In this article, we present a comprehensive review of research toward robust pattern recognition from the perspective of breaking three basic and implicit assumptions: closed-world assumption, independent and identically distributed assumption, and clean and big data assumption, which form the foundation of most pattern recognition models. Actually, our brain is robust at learning concepts continually and incrementally, in complex, open, and changing environments, with different contexts, modalities, and tasks, by showing only a few examples, under weak or noisy supervision. These are the major differences between human intelligence and machine intelligence, which are closely related to the above three assumptions. After witnessing the significant progress in accuracy improvement nowadays, this review paper will enable us to analyze the shortcomings and limitations of current methods and identify future research directions for robust pattern recognition.

In multiclass classification, convolutional neural network (CNN) is generally coupled with the cross-entropy (CE) loss, which only penalizes the predicted probability corresponding to a ground truth class and ignores the interclass relationship. We argue that CNN can be improved by using a better loss function. On the other hand, the Wasserstein distance (WD) is a well-known metric used to measure the distance between two distributions. Directly solving the WD problem requires a prohibitively large amount of computation time, whereas the cheaper iterative algorithms have a variety of shortcomings such as computational instability and difficulty in selecting parameters. In this paper, we address these issues by giving an analytical solution to the WD problem-for the first time, we find that for two distributions in hierarchically organized data space, WD has a closed-form solution, which we call “hierarchical WD (HWD).” We use this theory to construct novel loss functions that overcome the shortcomings of CE loss. To this end, multi-CNN information fusion that provides the basis for building category hierarchies is carried out first. Then, the semantic relationship among classes is modeled as a binary tree. Then, CNN coupled with an HWD-based loss, i.e., hierarchical Wasserstein CNN (HW-CNN), is trained to learn deep features. In this way, prior knowledge about the interclass relationship is embedded into HW-CNN, and information from several CNNs provides guidance in the process of training individual HW-CNNs. We conducted extensive experiments over two publicly available remote sensing data sets and achieved a state-of-the-art performance in scene classification tasks.

The effective retrieval of information from scanned handwritten documents is becoming essential with the increasing amounts of digitized documents. Therefore, developing efficient means of analyzing and recognizing these documents is of significant interest. Among these methods is word spotting, which has recently become an active research area. Different ensemble classifiers have been successfully proposed to improve the performance of a pattern recognition or a word spotting system. In this paper, we propose an enhanced internal structure of the Arabic handwritten word spotting hierarchical classifier. In addition, we propose two ensemble classifier combination methods to improve the performance of closed lexicon word spotting systems. These methods are, 1) the improved score word matching method, and 2) score evaluation method. Both methods calculate a new score by utilizing the confidence values (scores) given by the combined classifiers. Support Vector Machines (SVM) and Regularized Discriminant Analysis (RDA) have been utilized to implement the proposed ensemble classifier. The proposed methods have been tested using the CENPARMI Arabic handwritten documents database, and the results show that combining classifiers has a significant improvement on word spotting systems. The precision rate increased by 4% and 17% respectively, when the improved score matching method and the score evaluation method have been used.

This paper is concerned with learning from predictions. Predictions are obtained by ensemble of classifiers such as random forests (RF) or extra-trees. One assumes that estimators are semi independent so that they can be considered as prediction space. Hence we project our feature vector to the space of estimators obtaining responses from each of them. The responses for RFs are conditional class probabilities. The responses might be considered as projections onto some direction in quasi-orthogonal space which are decision trees of a RF. After that one creates the connected Riemannian manifold by computing a matrix of pairwise products of predictions for all trees in the RF. These matrices are symmetric and positive definite which is a necessary and sufficient condition to have a connected Riemannian manifold (R manifold). Because outputs of trees are conditional probabilities we have to create as many such matrices as there are classes. Stacking all these matrices together we obtain a tensor which is passed to Convolutional Neural Networks (CNN) for learning. We tested our algorithm on 11 datasets from UCI repository representing difficult classification problems. The results show very fast learning and convergence of loss and prediction accuracy. The proposed algorithm outperforms feature-based classical classifier ensembles (RFs and extra-trees) for every tested dataset from UCI repository.

Our concern is nonlinear clustering on large-scale dataset. While existing popular kernels (RBF, Polynomials, Spatial Pyramid, etc.) are popularly used for implicitly mapping data into a high-dimensional or infinite dimensional space in order to generalise linear clustering methods, using these kernels cannot make kernel clustering approaches directly applicable for large scale dataset, since large scale kernel matrix or similarity matrix consumes a lot of memory (e.g., 7,450 GB memory over 1 million samples of data). To solve this problem, we introduce an Euler clustering approach. Euler clustering employs Euler kernels in order to intrinsically map the input data onto a complex space of the same dimension as the input or twice, so that Euler clustering can get rid of kernel trick and does not need to rely on any approximation or random sampling on kernel function/matrix, whilst performing a more robust nonlinear clustering against noise and outliers. Moreover, since the original Euler kernel cannot generate a non-negative similarity matrix and thus is inapplicable to spectral clustering, we introduce a positive Euler kernel, and more importantly we have proved when it can generate a non-negative similarity matrix. We apply Euler kernel and the proposed positive Euler kernel to kernel k-means and spectral clustering so as to develop Euler k-means and Euler spectral clustering, respectively. An efficient Stiefel-manifold-based gradient method and an equivalent weighted positive Euler k-means are derived for fast computation of Euler spectral clustering and further alleviating the impact of discretization of the cluster membership indicators in Euler spectral clustering. The results show that the proposed Euler clustering approach achieves overall better clustering performance compared to using popular Mercer kernels and approximation models, whilst keeping the computational complexity of the same magnitude as the most popular linear clustering method k-means.

Graphology is a technique used to assess the writer's personality traits from his/her handwriting features. Manual feature extraction and analysis is a time consuming and labor intensive task. Therefore, computerized graphology systems have been developed by researchers to overcome these issues. In this paper, we present the latest state-of-the-art on computerized graphology systems.

Although widely used, Multilinear PCA (MPCA), one of the leading multilinear analysis methods, still suffers from four major drawbacks. First, it is very sensitive to outliers and noise. Second, it is unable to cope with missing values. Third, it is computationally expensive since MPCA deals with large multi-dimensional datasets. Finally, it is unable to maintain the local geometrical structures due to the averaging process. This paper proposes a novel approach named Compressed Submanifold Multifactor Analysis (CSMA) to solve the four problems mentioned above. Our approach can deal with the problem of missing values and outliers via SVD-L1. The Random Projection method is used to obtain the fast low-rank approximation of a given multifactor dataset. In addition, it is able to preserve the geometry of the original data. Our CSMA method can be used efficiently for multiple purposes, e.g. noise and outlier removal, estimation of missing values, biometric applications. We show that CSMA method can achieve good results and is very efficient in the inpainting problem as compared to [1], [2]. Our method also achieves higher face recognition rates compared to LRTC, SPMA, MPCA and some other methods, i.e. PCA, LDA and LPP, on three challenging face databases, i.e. CMU-MPIE, CMU-PIE and Extended YALE-B.

We propose a new approach to detect fake coins using their images in this paper. A coin image is represented in the dissimilarity space, which is a vector space constructed by comparing the image with a set of prototypes. Each dimension measures the dissimilarity between the image under consideration and a prototype. In order to obtain the dissimilarity between two coin images, the local keypoints on each image are detected and described. Based on the characteristics of the coin, the matched keypoints between the two images can be identified in an efficient manner. A post-processing procedure is further proposed to remove mismatched keypoints. Due to the limited number of fake coins in real life, one-class learning is conducted for fake coin detection, so only genuine coins are needed to train the classifier. Extensive experiments have been carried out to evaluate the proposed approach on different data sets. The impressive results have demonstrated its validity and effectiveness.

There are many difficulties facing a handwritten Arabic recognition system such as unlimited variation in human handwriting, similarities of distinct character shapes, interconnections of neighboring characters and their position in the word. This paper presents a handwritten Arabic character recognition system based on BA algorithm. BA algorithm is adopted to reduce the feature set size and to improve the accuracy rate. The proposed system is trained and tested by four well-known classifiers; Bayes Network (BN), artificial neural network (ANN), K-nearest neighbors (KNN), and Random forest (RF) with CENPARMI dataset. The proposed optimization algorithm obtained promising results in terms of classification accuracy as the proposed system is able to recognize 91.59 % of our test set correctly, as well as in terms of computational time reduction. BA algorithm is more efficient in most experiments when comparing with GA and PSO. When compared our results with other related works we find that our result is the highest among other published results.

This paper presents a new flexible approach to predict the gender of the writers from their handwriting samples. Handwriting features can be extracted from different methods. Therefore, the multi-feature sets are irrelevant and redundant. The conflict of the features exists in the sets, which affects the accuracy of classification and the computing cost. This paper proposes a Mutual Information (MI) approach, that focuses on feature selection. The approach can decrease redundancies and conflicts. In addition, it extracts an optimal subset of features from the writing samples produced by male and female writers. The classification is carried out using a Support Vector Machine (SVM) on two databases. The first database comes from the ICDAR 2013 competition on gender prediction, the other database contains the Registration-Document-Form (RDF) database in Chinese. The proposed and compared methods were evaluated on both databases. Results from the methods highlight the importance of feature selection for gender prediction from handwriting.

In this paper, we initiate a theoretical study on N-expert fusion (N ≥ 2) in the context of biometric authentication (BA). Optimal fusion weights, which depend on performances and variances of, and correlations among individual base-experts have been found, and we also give and prove some new theorems that serve as the basis for analyzing the performance of the overall system. Our conclusion is that provided that optimal weights are used as fusion coefficients, linear fusion will definitely lead to a better performance than the best individual expert. This contradicts many existing conclusions, which assert that fusion is not always beneficial and that performance improvement due to fusion is guaranteed only when some conditions as to baseexperts' performances, variances, and correlations are satisfied. Besides, for the first time the definition of correlation in the context of BA is clearly and explicitly given to avoid the longstanding ambiguity and vagueness concerning this term, and we make an initial attempt to propose and investigate three types of correlation coefficients. Furthermore, the connection between our proposed optimal fusion method and Fisher's discriminant is discussed. Extensive experiments have been conducted to confirm our theoretical results and construct counter-examples for the existing conclusions.

We propose a variable-length signature for near-duplicate image matching in this paper. An image is represented by a signature, the length of which varies with respect to the number of patches in the image. A new visual descriptor, viz., probabilistic center-symmetric local binary pattern, is proposed to characterize the appearance of each image patch. Beyond each individual patch, the spatial relationships among the patches are captured. In order to compute the similarity between two images, we utilize the earth mover's distance which is good at handling variable-length signatures. The proposed image signature is evaluated in two different applications, i.e., near-duplicate document image retrieval and near-duplicate natural image detection. The promising experimental results demonstrate the validity and effectiveness of the proposed variable-length signature.

This paper introduces new handwritten databases of selected words in the five Middle-Eastern languages of Arabic, Dari, Farsi, Pashto and Urdu. The databases share a common lexicon of forty words that are related to finance and are used in daily life. The five databases have been collected from over 1600 native writers located in four countries. Recognition results for each of the databases are also presented. Results come from three classifiers (Support Vector Machines, Modified Quadratic Discriminant Function. And Multi-layer Perceptron) which were implemented for recognition of the words based on gradient features. Given the diversity of the data, the results demonstrate the effectiveness of the implemented process in learning and recognizing samples of handwritten words from different languages. In addition, full page handwritten documents of each language are presented, with approximately forty pages per language. Each document has associated ground truth information.

A new near-duplicate document image matching approach is proposed. Globally, we model the spatial arrangements of objects in an image. Locally, the micro-patterns within each object are captured. To define a micro-pattern, the N-nary center-symmetric gray value differences in an image local neighborhood of a variable radius are exploited. A visual descriptor is proposed to characterize the appearance of the object based on micro-pattern distributions. By combining the global and local features, each document image is represented by a compact signature with a variable length. We employ Earth Mover's Distance for image dissimilarity computation, which stands out for its remarkable ability to tolerate the instability of object segmentation by allowing many-to-many correspondence among objects. Extensive experiments on two data sets demonstrate the effectiveness of the proposed approach.

In the field of biometric authentication, it is a promising trend to perform score fusion to improve authentication accuracy. Many empirical studies have shown the effectiveness of score fusion; however, some other researchers assert that fusion is not always beneficial. Despite considerable empirical efforts, to the best of our knowledge, the research devoted to the theoretical analysis of fusion can be found only in the paper by Poh and Bengio published in 2005. Unfortunately, we find that the variance reduction-equal error rate (VR-EER) model, which is the theoretical basis of this reference, is incorrect and the resulting conclusions are arguable. Besides, we find that the conclusions from several other empirical studies are arguable too. In this paper, using Fermat's theorem and the connection between F-ratio and EER, we conduct a systematic theoretical study on how correlation and performances of base-experts affect fusion, giving the underlying reason why VR-EER model and the above conclusions are wrong. Contrary to these existing conclusions, we prove that provided fusion weights are selected according to our proposed criterion, the combined system will definitely be superior to all the base-experts, regardless of correlation, performances, or variances of base-experts. Experiments are carried out to validate the conclusions of ours and construct counter-examples for the existing conclusions.

The study of RMB (renminbi bank note, the paper currency used in China) serial number recognition draws more and more attention in recent years, for reducing financial crime, improving financial market stability and social security. The accuracy of RMB recognition relies heavily on the extraction, which is a challenging problem due to background variations and uneven illumination. In this paper, we present a new system that extracts the RMB characters directly from scanned RMB images. First, two different techniques, namely skew correction and orientation identification are used to detect the region which contains RMB serial number. Then the detected text region is binarized by a combined thresholding technique. After that, a local contrast average method is introduced to extract the RMB characters from the binarization result. The experiments demonstrate that the proposed binarization method outperforms other well-known methods. For character extraction, we report an overlap-recall rate of 79.68% and an overlap-precision rate of 98.10% respectively.

Facing high error rates and slow recognition speed for full text transcription of unconstrained handwriting images, keyword spotting is a promising alternative to locate specific search terms within scanned document images. We have previously proposed a learning-based method for keyword spotting using character hidden Markov models that showed a high performance when compared with traditional template image matching. In the lexicon-free approach pursued, only the text appearance was taken into account for recognition. In this paper, we integrate character n-gram language models into the spotting system in order to provide an additional language context. On the modern IAM database as well as the historical George Washington database, we demonstrate that character language models significantly improve the spotting performance.

Large amounts of handwritten documents have been digitized, and the need to search and index these documents is increasing to make them more accessible. Different word spotting systems have been proposed to search for words for this purpose. Since the precision of the word spotting system is crucial, verifying the results of a word spotting system is becoming an effective approach to improve the system performance. In this paper, we propose two verification models for Arabic word spotting systems. Both models make use of a holistic classifier. The first model is based on matching the results of the word spotting system with those of the holistic classifier, while the other model derives a new score evaluation based on the two results. Verifying a word spotting system using these models can significantly improve its performance, since the precision rate increased from 74% to 77.7% and 84.4% respectively with the Word Matching and Score Evaluation models of verification, at 50% recall.

The issue of near duplicate document image retrieval is addressed in this paper, which is characterized by not only encoding each individual word in the image but also modeling its local spatial configuration. On representing each word in the image as a string in terms of its shape characteristics, a lexicon is first learnt from a training set. Then a word in an arbitrary document image can be soft assigned to a weighted combination of several nearest neighbors in the lexicon. The rationale behind soft-assignment is to tolerate the distortions induced by character segmentations which are error-prone in degraded document images. Most importantly, we look beyond the single word and capture the local spatial configuration for each word which plays a very important role in human perception. It provides much useful information in discriminating between different document images compared with the single word. A graph, benefitting from its great representative power, is built for each word to model its relationships with the neighborhoods locally. The local word spatial configurations are integrated within the inverted file index structure to achieve scalable retrieval. Thus the retrieval of near duplicate document images is formulated as a voting problem. Experimental results on 45,000 document images demonstrate that the proposed approach brings significant improvements in successful retrieval of near duplicate images.

The affinity propagation (AP) clustering algorithm has received much attention in the past few years. AP is appealing because it is efficient, insensitive to initialization, and it produces clusters at a lower error rate than other exemplar-based methods. However, its single-exemplar model becomes inadequate when applied to model multisubclasses in some situations such as scene analysis and character recognition. To remedy this deficiency, we have extended the single-exemplar model to a multi-exemplar one to create a new multi-exemplar affinity propagation (MEAP) algorithm. This new model automatically determines the number of exemplars in each cluster associated with a super exemplar to approximate the subclasses in the category. Solving the model is NP-hard and we tackle it with the max-sum belief propagation to produce neighborhood maximum clusters, with no need to specify beforehand the number of clusters, multi-exemplars, and superexemplars. Also, utilizing the sparsity in the data, we are able to reduce substantially the computational time and storage. Experimental studies have shown MEAP's significant improvements over other algorithms on unsupervised image categorization and the clustering of handwritten digits.

This paper presents a robust method for handwritten text line extraction. We use morphological dilation with a dynamic adaptive mask for line extraction. Line separation occurs because of the repulsion and attraction between connected components. The characteristics of the Arabic script are considered to ensure a high performance of the algorithm. Our method is evaluated on the CENPARMI Arabic handwritten documents database which contains multi-skewed and touching lines. With a matching score of 0.95, our method achieved precision and recall rates of 96:3% and 96:7% respectively, which demonstrate the effectiveness of our approach.

We present a statistical hypothesis testing method for handwritten word segmentation algorithms. Our proposed method can be used along with any word segmentation algorithm in order to detect over-segmented or under-segmented errors or to adapt the word segmentation algorithm to new data in an unsupervised manner. The main idea behind the proposed approach is to learn the geometrical distribution of words within a sentence using a Markov chain or a Hidden Markov Model (HMM). In the former, we assume all the necessary information is observable, where in the latter, we assume the minimum observable variables are the bounding boxes of the words, and the hidden variables are the part of speech information. Our experimental results on a benchmark database show that not only we can achieve a lower over-segmentation and under-segmentation error rate, but also a higher correct segmentation rate as a result of the proposed hypothesis testing.

With the ever-increasing amounts of published materials being made available, developing efficient means of locating target items has become a subject of significant interest. Among the approaches adopted for this purpose is word spotting, which enables the identification of documents through the use of pertinent keywords. This paper reports on an effective method of word spotting for Arabic handwritten documents that takes into consideration the nature of Arabic handwriting. Parts of Arabic Words (PAWs) form the basic components of this search process, and a hierarchical classifier (consisting of a set of classifiers each trained on a different part of the input pattern) is implemented. For the first time in Arabic word spotting, language models are incorporated into the process of reconstructing words from PAWs. Details of the method and promising experimental results are also presented.

In recognition of Off-line handwritten characters and signatures, stroke extraction is often a crucial step. Given the large number of Chinese handwritten characters, pattern matching based on structural decomposition and analysis is useful and essential to Off-line Chinese recognition to reduce ambiguity. Two challenging problems for stroke extraction are: 1) how to extract primary strokes and 2) how to solve the segmentation ambiguities at intersection points. In this paper, we introduce a novel approach based on Optimum Paths(AOP) to solve this problem. Optimum Paths(AOP) are derived from the degree information and continuation property, we use them to tackle these two problems. Compared with other methods, the proposed approach has extracted strokes from Off-line Chinese handwritten characters with better performance.

Matching pedestrians across disjoint camera views is a challenging task, since their observations are separated in time and space and their appearances may vary considerably. Recently, some approaches of matching pedestrians have been proposed. However, these approaches either used too complex representations or only considered the color information and discarded the spatial structural information of the pedestrian. In order to describe the spatial structural information in color space, we propose a distance-based local binary pattern (DLBP) descriptor. Besides the spatial structural information, the color itself namely its intensity value is also an important feature in matching pedestrians across disjoint camera views. In order to effectively combine these two kinds of information, we further propose a novel CI_DLBP descriptor, which unifies the color intensity and DLBP by learning the joint distributions (2-D histograms) of the DLBP and color intensity at each channel. In addition, different from the previous approaches in which the pedestrians matching is based on their whole bodies, we develop a part-based pedestrian representation because the color density and spatial structural information between the upper outer garment and the lower garment worn by the pedestrian is usually different. Experimental results on challenging realistic scenarios and VIPeR dataset validate the proposed DLBP operator, the CI_DLBP descriptor, and the part-based pedestrian representation for pedestrian matching across disjoint camera views. Compared with existing methods based on color information, this new CI_DLBP approach performs better.

This paper introduces an advanced age-determination technique using hybrid facial features and Kernel Spectral Regression, a nonlinear dimensionality reduction method. In the preprocessing stage, the logarithmic nonsubsampled contourlet transform (NSCT) is conducted to denoise and amplify facial wrinkles that help to distinguish young faces from elder ones. Then the hybrid facial features that combine both local and holistic features are extracted from the preprocessed images. Our novel Uniform Local Ternary Patterns (ULTP) are used as the local features. Meanwhile the holistic features are extracted by using the Active Appearance Model (AAM) to encode each face. Kernel Spectral Regression is used to minimize inter-class distances while maximizing intra-class distances of feature sets. These reduced features are used to classify faces into two age groups (age-classification). An age-determination function is then constructed for each age group in accordance with physiological growth periods for humans - pre-adult (youth) and adult. Compared to published results, this method yields promising results in overall mean absolute error (MAE), mean absolute error per decade of life (MAE/D), and cumulative match score in various face aging corpuses.

In this paper, we will present a novel framework of utilizing periocular region for age invariant face recognition. To obtain age invariant features, we first perform preprocessing schemes, such as pose correction, illumination and periocular region normalization. And then we apply robust Walsh-Hadamard transform encoded local binary patterns (WLBP) on preprocessed periocular region only. We find the WLBP feature on periocular region maintains consistency of the same individual across ages. Finally, we use unsupervised discriminant projection (UDP) to build subspaces on WLBP featured periocular images and gain 100% rank-1 identification rate and 98% verification rate at 0.1% false accept rate on the entire FG-NET database. Compared to published results, our proposed approach yields the best recognition and identification results.

In this paper we propose a novel Contourlet Appearance Model (CAM) that is more accurate and faster at localizing facial landmarks than Active Appearance Models (AAMs). Our CAM also has the ability to not only extract holistic texture information, as AAMs do, but can also extract local texture information using the Nonsubsampled Contourlet Transform (NSCT). We demonstrate the efficiency of our method by applying it to the problem of facial age estimation. Compared to previously published age estimation techniques, our approach yields more accurate results when tested on various face aging databases.

A vast number of historical and badly degraded document images can be found in libraries, public, and national archives. Due to the complex nature of different artifacts, such poor quality documents are hard to read and to process. In this paper, a novel adaptive binarization algorithm using ternary entropy-based approach is proposed. Given an input image, the contrast of intensity is first estimated by a grayscale morphological closing operator. A double-threshold is generated by our Shannon entropy-based ternarizing method to classify pixels into text, near-text, and non-text regions. The pixels in the second region are relabeled by the local mean and the standard deviation. Our proposed method classifies noise into two categories which are processed by binary morphological operators, shrink and swell filters, and graph searching strategy. The method is tested with three databases that have been used in the Document Image Binarization Contest 2009 (DIBCO 2009), the Handwriting Document Image Binarization Contest 2010 (H-DBCIO 2010), and the International Conference on Frontier in Handwriting Recognition 2010 (ICFHR 2010). The evaluation is based upon nine distinct measures. Experimental results show that our proposed algorithm outperforms other state-of-the-art methods.

A graph matching approach is proposed to retrieve envelope images from a large image database. First, the graph representation of an envelop image is generated based on the image segmentation results, in which each node corresponds to one segmented region. The attributes of nodes and edges in the graph are described by characteristics of the envelope image. Second, a minimum weighted bipartite graph matching method is employed to compute the distance between two graphs. Finally, the whole retrieval system including two principal stages is presented, namely, rough matching and fine matching. The experiments on a database of envelope images captured from real-life mail pieces demonstrate that the proposed method achieves promising results.

In order to spot the digits in a handwritten document, each component is sent to a classifier. This is a time consuming process because a document usually contains several hundred components. A method is presented to reduce the number of candidate components from a handwritten document sent to the classifier. Furthermore, since the classifier does not contain a rejection class, this led to several misclassifications. To lessen this, a verification post processing module was developed in order to reject some false positives. We reached an overall precision of 80% and 83.3% recall on our test set of handwritten documents.

Advances in digital technology have greatly facilitated the design of new type fonts. Today, hundreds of thousands of fonts can be found in various visual appearances or styles, which are used in digital publishing and information display. As a result, it has become important to find ways of evaluating their impact on our daily lives: (1) ease in reading, (2) comprehension of the texts, and (3) eye-strain. This paper summarizes an in-depth inquiry into the following topics: (a) impact of fonts on digital publishing and display, (b) the influence of typographic features on reading, (c) the role of fonts in reading, (d) effect of spacing on reading speed and comprehension, and (e) machine reading of early styles of ancient Chinese characters. Several insightful questions on this subject are asked, and answers have been provided through this paper and the oral presentations. A comprehensive list of references is included at the end of each section for further studies and research.

More and more fonts have sprung up in recent years in digital publishing industry and reading devices. In this paper, we focus on methods of evaluating digital Chinese fonts and their typeface characteristics to seek a good way to enhance the character recognition rate. To accomplish this, we combined psychological analysis methods with statistical analysis. It involved an extensive survey of distinctive features of eighteen popular digital typefaces. Survey results were tabulated and analyzed statistically. Then another objective experiment was conducted using the best six fonts derived from the survey results. These experimental results reveal an effective way of choosing legible digital fonts most suitable for comfortable reading of books, magazines, newspapers, and for display of texts on cell-phones, e-books, and digital libraries, and finding out the features for improving character legibility of different Chinese typefaces. The relationships among legibility, eye-strain, and myopia, will be discussed.

A face image can be represented by a combination of large-and small-scale features. It is well-known that the variations of illumination mainly affect the large-scale features (low-frequency components), and not so much the small-scale features. Therefore, in relevant existing methods only the small-scale features are extracted as illumination-invariant features for face recognition, while the large-scale intrinsic features are always ignored. In this paper, we argue that both large-and small-scale features of a face image are important for face restoration and recognition. Moreover, we suggest that illumination normalization should be performed mainly on the large-scale features of a face image rather than on the original face image. A novel method of normalizing both the Small-and Large-scale (S&L) features of a face image is proposed. In this method, a single face image is first decomposed into large-and small-scale features. After that, illumination normalization is mainly performed on the large-scale features, and only a minor correction is made on the small-scale features. Finally, a normalized face image is generated by combining the processed large-and small-scale features. In addition, an optional visual compensation step is suggested for improving the visual quality of the normalized image. Experiments on CMU-PIE, Extended Yale B, and FRGC 2.0 face databases show that by using the proposed method significantly better recognition performance and visual results can be obtained as compared to related state-of-the-art methods.

Most research on fonts is related to legibility, readability and recognition. There are only a few studies on typefaces and their potential personality traits. In this paper, we focus on the visual expression of typefaces and their design characteristics. The relationship between typefaces and their personality traits is investigated. By using statistical analyses on data collected from participants who filled out a survey, the correlation between fonts and personality traits is explored. Fonts used within this study are grouped according to their personality traits. The typical typographical and aesthetic characteristics of typefaces in these groups are examined in detail.

This paper presents an advanced age-determination technique that combines holistic and local features derived from an image of the face. A 30×1 Active Appearance Model (AAM) linear encoding of each face is produced to work as holistic features. Meanwhile, local features are extracted by using Local Ternary Patterns (LTP). These combined features are used to classify faces into one of two age groups (age-classification). An age-determination function is then constructed for each age group in accordance with physiological growth periods for humans - pre-adult (youth) and adult. Compared to published results, this method yields the highest accuracy rates in overall mean absolute error (MAE), mean absolute error per decade of life (MAE/D), and cumulative match score.

In practical applications, errors should not be treated equally, but conditionally. In this paper, errors are categorized based on different costs in misclassification. Accordingly, the characteristics of the error categorization and the corresponding strategies for correcting them are proposed. Verification based on Arabic Handwritten Numeral Recognition is considered as one application to utilize these definitions and strategies. As a result, the recognition results improved from 98.47% to 99.05%, and errors were significantly reduced by over 35% compared to previous studies. When a rejection measurement was applied, and the rejection threshold was adjusted to maintain the same error rate, both the recognition rate and reliability increased from 96.98% to 97.89% and from 99.08% to 99.28%, respectively.

In this paper, we present an approach for word spotting in Gray-scale Pashto Documents, written in modified Arabic scripts. Various profile and transitional features are extracted from gray-scale word images. The gray-scale feature vectors are then converted into binary feature vectors by replacing each value within the gray-scale feature vectors with its binary equivalents. In this way, we have enabled the alignment of the gray-scale feature vectors via a faster binary pattern matching algorithm, i.e., Correlation Similarity Measure (CORR). The approach has effectively handled the handwriting variations of 200 different writers. The average precision rate achieved is 94.75 % for an average recall of 60.25%. The time taken for matching every set of two word images is 1.43 ms.

We process the unideal iris images that are acquired in an unconstrained situation and are affected severely by gaze deviations, eyelid and eyelash occlusions, non uniform intensities, motion blurs, reflections, etc. The proposed unideal iris recognition algorithm has two novelties as compared to the previous works; firstly, we propose to deploy a region-based active contour model to segment an unideal iris image with intensity inhomogeneity; Secondly, an iterative algorithm, called the Modified Contribution- Selection Algorithm (MCSA), is used in the context of coalitional game theory to select a subset of informative features without compromising the recognition rate. The verification performance of the proposed scheme is validated using the UBIRIS Version 1, the ICE 2005, and the WVU Unideal datasets.

This paper views target detection and pattern recognition as a kind of communications problem and applies error-correcting coding to the outputs of a convolutional neural network to improve the accuracy and reliability of detection and recognition of targets. The outputs of the convolutional neural network are designed according to codewords with maximum Hamming distances. The effects of the codewords on the performance of the convolutional neural network in target detection and recognition are then investigated. Images of hand-written digits and printed English letters and symbols are used in the experiments. Results show that error-correcting output coding provides the neural network with more reliable decision rules and enables it to perform more accurate and reliable detection and recognition of targets. Moreover, our error-correcting output coding can reduce the number of neurons required, which is highly desirable in efficient implementations.

In handwriting recognition, confusing/conflicting writing styles can result in irreducible errors, so the study of writing style consistencies is important for applications. In Arabic Handwritten Numeral Recognition, most errors occur between samples of classes two and three due to their very similar shapes in some writing styles. In this paper, an automated writing style detection process is effectively implemented in the pair-wise verification of samples in these two classes. As a result, the recognition results have improved significantly with a reduction by 25% of previous errors. With rejection, when the LDA (Linear Discriminant Analysis) measurement rejection threshold is adjusted to maintain the same error rate, the recognition rate increases from 96.87% to 97.81%.

Robust localization of inner/outer boundary from an iris image plays an important role in iris recognition. However, the conventional iris/pupil localization methods using the region-based segmentation or the gradient-based boundary finding are often hampered by non-linear deformations, pupil dilations, head rotations, motion blurs, reflections, non-uniform intensities, low image contrast, camera angles and diffusions, and presence of eyelids and eyelashes. The novelty of this research effort is that we apply a parallel game-theoretic decision making procedure by using the modified Chakra borty and Duncan's algorithm, which integrates the region-based segmentation and gradient-based boundary finding methods and fuses the complementary strengths of each of these individual methods. This integrated scheme forms a unified approach, which is robust to noise and poor localization.

Since the Urdu language has more isolated letters than Arabic and Farsi, a research on Urdu handwritten word is desired. This is a novel approach to use the compound features and a Support Vector Machine (SVM) in offline Urdu word recognition. Due to the cursive style in Urdu, a classification using a holistic approach is adapted efficiently. Compound feature sets, which involves in structural and gradient features (directional features), are extracted on each Urdu word. Experiments have been conducted on the CENPARMI Urdu Words Database, and a high recognition accuracy of 97.00% has been achieved.

We propose a novel word spotting system for Urdu words within handwritten text lines. Spatial information of diacritics is integrated to the detection of the main connected components in candidate words generation. An Urdu word recognition system is effectively designed and applied to classify the candidate words. In this word recognition system, compound features and SVM were adapted. The verification/rejection process was based on the outputs from the Urdu word recognition system and the image's global features were applied to achieve a promising result. As a result, a high 92.11% correct segmentation rate, a 50.75% word spotting precision rate were achieved while maintaining a 70.1% recall on CENPARMI's Urdu Database.

Document Script Identification (DSI) is a very useful application in document processing. This paper presents a method for this application that uses a new noise tolerant feature, the Downgraded Pixel Density feature. Compared to other features widely used in existing DSI solutions, this new feature is much more robust to variations in slant, font and style of printed documents. Experimental results show that the method achieves promising identification performances.

In this paper, we propose a novel illumination-normalization method. By using the combination of the Kernel Principal Component Analysis (KPCA) and Pre-image technology, this method can restore the frontal-illuminated face image from a single non-frontal-illuminated face image. In this method, a frontal-illumination subspace is first learned by KPCA. For each input face image, we project its large-scale features, which are affected by illumination variations, onto this subspace to normalize the illumination. Then the frontal-illuminated face image is reconstructed by combining the small- and the normalized large- scale features. Unlike most existing techniques, the proposed method does not require any shape modeling or lighting estimation. As a holistic reconstruction, KPCA+Pre-image technology incurs less local distortion. Compared to directly applying KPCA+Pre-image technology on the original image, our proposed method can be better at processing an image of a face that is outside the training set. Experiments on CMU-PIE and Extended Yale B face databases show that the proposed method outperforms state-of-the-art algorithms.

A novel and efficient method for correction of slant angles in handwritten numeral strings is proposed. For the first time, the statistical distribution of slant angles in handwritten numerals is investigated and the effects of slant correction on the segmentation of handwritten numeral strings are shown. In our proposed slant correction method, utilizing geometric features, a Component Slant Angle (CSA) is estimated for each connected component independently. A weighted average is then used to compute the String Slant Angle (SSA), which is applied uniformly to correct the slant of all the components in numeral strings. Our experimental results have revealed novel statistics for slant angles of handwritten numeral strings, and also showed that slant correction can significantly improve extraction of segmentation features and segmentation accuracy of touching numerals. Comparison between our slant correction algorithm and similar algorithms in the literature show that our algorithm is more efficient, and on average it has a faster running time.

To carry out face transformation, this paper presents new numerical algorithms, which consist of two parts, namely, the harmonic models for changes of face characteristics and the splitting techniques for grayness transition. The main method in this paper is a combination of the finite-volume method (FVM) with Delaunay triangulation to solve the Laplace equations in the harmonic transformation of face images. The advantages of the FVM with Delaunay triangulation are given as follows: 1) easy to formulate the linear algebraic equations; 2) good in retaining the pertinent geometric and physical need; and 3) less central processing unit time needed. Numerical and graphical experiments have been conducted for the face transformation from a female (woman) to a male (man), and vice versa. The computed sequential errors are O(N⁻³/²), where N² is the division number of a pixel into subpixels. These computed errors coincide with the analysis on the splitting-shooting method (SSM) with piecewise constant interpolation in the previous paper of Li and Bai. In computation, the average absolute errors of restored pixel grayness can be smaller than 2 out of 256 grayness levels. The FVM is as simple as the finite-difference method (FDM) and as flexible as the finite-element method (FEM). Hence, the FVM is particularly useful when dealing with large face images with a huge number of pixels in shape distortion. The numerical transformation of face images in this paper can be used not only in pattern recognition but also in resampling, image morphing, and computer animation.

It is known that convolutional neural networks (CNNs) are efficient for optical character recognition (OCR) and many other visual classification tasks. This paper applies error-correcting output coding (ECOC) to the CNN for segmentation-free OCR such that: 1) the CNN target outputs are designed according to code words of length N; 2) the minimum Hamming distance of the code words is designed to be as large as possible given N. ECOC provides the CNN with the ability to reject or correct output errors to reduce character insertions and substitutions in the recognized text. Also, using code words instead of letter images as the CNN target outputs makes it possible to construct an OCR for a new language without designing the letter images as the target outputs. Experiments on the recognition of English letters, 10 digits, and some special characters show the effectiveness of ECOC in reducing insertions and substitutions.

This paper presents a linear discriminant analysis based measurement (LDAM) on the output from classifiers as a criterion to reject the patterns which cannot be classified with high reliability. This is important in applications (such as in processing of financial documents) where errors can be very costly and therefore less tolerable than rejections. To implement the rejection, which can be considered to be a two-class problem of accepting the classification result or otherwise, Linear discriminant analysis (LDA) is used to determine the rejection threshold at a new approach. LDAM is designed to take into consideration the confidence values of the classifier outputs & the relations between them, and it is an improvement over traditional rejection measurements such as first rank measurement (FRM) and first two ranks measurement (FTRM). Experiments are conducted on the CENPARMI Arabic isolated numerals database. The results show that LDAM is more effective, and it can achieve a higher reliability while achieving a high recognition rate.

A new isolated handwritten Farsi numeral recognition algorithm is proposed in this paper, which exploits the sparse and over-complete structure from the handwritten Farsi numeral data. In this research, the sparse structure is represented as an over-complete dictionary, which is learned by the K-SVD algorithm. These atoms in this dictionary are adopted to initialize the first layer of the convolutional neural network (CNN), the latter is then trained to do the classification task. Data distortion techniques are also applied to promote the generalization capability of the trained classifier. Experiments have shown that good results have been achieved in CENPARMI handwritten Farsi numeral database.

Transforming handwriting into digital text and recognition of handwritten patterns opens a vast scope of application opportunities from searching for handwritten notes and document management to causing actions by writing symbols. Despite receiving a great attention, a massive number of applications, and a huge research effort, recognition of handwritten text has not still reached a desired efficiency and is an active area of research. One of the most important factors that makes handwriting recognition a challenging task is the huge variety of writing styles which can not be captured efficiently through available classification methods using current feature descriptors. Our approach to gain performance in online character recognition is to design more representative features for handwritten character representation in order to tackle the huge inter-class variability problem and increase recognition accuracy. The representation can also be used in recognition of other online planar patterns. The experimental results show that proposed representation with SVM classifier outperforms best reported recognition rates for Arabic characters in a writer-independent system.

A new courtesy amount recognition module of CENPARMIpsilas check reading system (CRS) is proposed in this paper. The module consists of 3 main segments: pre-processing, segmentation and recognition, and post-processing. A new feedback-based segmentation algorithm is adopted for the segmentation task. Besides one individual numeral recognizer for numerals from dasia0psila to dasia9psila, one convolutional neural network(CNN) recognizer for ldquo00rdquo and ldquo000rdquo numeral strings is also integrated into our module for the recognition task. The experimental results on the Quebec Bell Check database show that the recognition rate of the courtesy amount has improved from 41.2% to 74.3%.

When linear support vector machines (SVMs) are applied to multi-class text categorization in industry, the size of the linear SVM model is very large, usually greater than several gigabytes. As a result, the model cannot directly fit into the computer memory and the classification process is slow. In this paper, a novel method based on vector norm is proposed to shrink the model size significantly without sacrificing the classification accuracy. Also, we propose a cache-efficient implementation of multi-class linear SVMs in the classification phase. Our experimental results have shown that on Yahoo-Korea dataset the proposed method can shrink the model size from 5.2 gigabytes to 260 megabytes and the efficient implementation of linear SVM has obtained a speedup factor of 44.

A sparse representation based method is proposed for text detection from scene images. We start with edge information extracted using Canny operator and then group these edge points into connected components. Each connected component is labeled as text or non-text by a two-level labeling process: pixel level labeling and connected component labeling. The core of the labeling process is a sparsity test using an over-complete dictionary, which is learned from edge segments of isolated character images. Layout analysis is further applied to verify these text candidates. Experimental results show that improvements in both recall rate and detection accuracy in text detection have been achieved.

This paper presents a multimodal approach for a biometrics verification system. It is based on face and hand images captured by a cell phone. The algorithm includes all parts that are required for face and hand verification, such as feature extraction, classification and authentication. To find local facial features, such as eyes, mouth and nose, we apply a point distribution model and active shape models. We use the same system to find distinctive points in hand geometry. The face feature vector is constructed by applying a Gabor filter to the image and extracting the key points found by an active shape model. The palm feature vector contains characteristics of the hand geometry features. A support vector machine (SVM) is applied to verify the identity of the user. One SVM machine is built for each person in the database to distinguish that person from others. To test the algorithm we built our own database containing face and hand images taken by a cell phone camera. The database contains 480 frontal face images and 120 hand images of 30 persons (16 face images and 4 hand images per person).

A novel text segmentation method from complex background is presented in this paper. The idea is inspired by the recent development in searching for the sparse signal representation among a family of over-complete atoms, which is called a dictionary. We assume that the image under investigation is composed of two components: the foreground text and the complex background. We further assume that the latter can be modeled as a piece-wise smooth function. Then we choose two dictionaries, where the first one gives sparse representation to one component and non-sparse representation to another while the second one does the opposite. By looking for the sparse representations in each dictionary, we can decompose the image into the two composing components. After that, text segmentation can be easily achieved by applying simple thresholding to the text component. Preliminary experiments show some promising results.

This paper presents a study on some approaches in face reconstruction and the methodology used. The results and drawbacks of each method are discussed. To construct the 3D face model there are some methods which select the features automatically or manually. First a method in textured 3D face reconstruction using two 2D images from any angle is discussed which does not need any particular database, but it has to define the feature points manually. Then we describe automatic 2D to 3D face reconstruction from a single frontal face image which needs the use of USF Human ID 3-D database. Afterwards we talk about a Model Based Face Reconstruction for Animation. We also describe briefly about 3D face modeling by fusing multiple 2D images which is fully automatic and via an EM approach which uses the shape and pose parameters. Finally we describe a Rapid Modeling of Animated Faces from Video method and 2D face reconstruction using a minimum set of feature points.

In this paper, we propose a new approach for face pose estimation based on two images from different views under certain conditions. Using a weak-perspective imaging model, six pose parameters were deduced, with four pairs of feature points properly chosen across the two face images. Through the scan-iteration algorithm, a robust performance was achieved, without solving non-linear equations. Comparing to some other methods which estimate rotation matrix based on fundamental matrix (F), our method focuses on the "absolute pose" with respect to front view rather than the "relative pose" between the two face images. "Absolute pose" is indispensable in most situations especially for 3D face modeling. Since our method does not depend on any 3D face models and frontal-view images, it can be applied not only to face recognition and 3D face modeling, but also to other relevant applications. Experimental results demonstrate the efficiency of our method

In this paper, a Genetic Binary Particle Swarm Optimization (GBPSO) model is proposed, and its performance is compared with the regular binary Particle Swarm Optimizer (PSO), introduced by Kennedy and Eberhart. In the original model, the size of the swarm was fixed. In our model, we introduce birth and death operations in order to make the population very dynamic. Since birth and mortality rates change naturally with time, our model allows oscillations in the size of the population. Compared to the original PSO model, and Genetic Algorithms, our strategy proposes a more natural simulation of the social behavior of intelligent animals. The experimental results show that compared to original PSO, our GBPSO model can reach broader domains in the search space and converge faster in very high dimensional and complex environments.

This paper presents two fast and robust algorithms for word skew and slant corrections based on Radon transform. For the skew correction, we maximize a global measure which is defined by Radon transform of image and its gradient to estimate the slope. For the slant correction, Radon transform is used to estimate the long strokes and a word slant is measured by the average angle of these long strokes. Compared with the previous methods, these two algorithms do not require the setting of parameters heuristically. Moreover, the algorithms perform well on words of short length, where the traditional methods usually fail.

Noise and complex backgrounds often make the thresholding of degraded document images difficult. In this paper, we propose a new threshold selection method to handle severely degraded document images. First, multiscale image description is adopted to analyze the image edge. From this, edge pixel pair information is derived and recorded by a graylevel co-occurrence matrix. An appropriate threshold value is chosen by measuring the edge pixel pair co-occurrence matrix. The new method is tested with degraded document images. The experimental results show it is resistant to noise and complex backgrounds.

Multi-channel Gabor filtering has been widely used in texture classification. In this paper, Gabor filters have been applied to the problem of script identification in printed documents. Our work is divided into two stages. Firstly, a Gabor filter bank is appropriately designed so that extracted rotation-invariant features can handle scripts that are similar in shape and even share many characters. Secondly, the steerability property of Gabor filters is exploited to reduce the high computation cost resulted from the frequent image filtering, which is a common problem encountered in Gabor filter related applications. Results from preliminary experiments are quite promising, where Chinese, Japanese, Korean and English are considered. Over 98.5 % language identification rate can be achieved while image filtering operations have been reduced by 40%.

Summary form only given. There has been much research on human thought processes, but little on the inference of such knowledge into the computer, for the purpose of handwriting recognition. Although modern recognition engines can recognize many handwritten symbols and cursive scripts with a high level of accuracy, they often make foolish or unreasonable mistakes. These engines often act like black boxes, which is why they make such mistakes on characters that would normally be easily recognized by human beings. To break through this level of accuracy, we have to look back and explore more human aspects, to better understand their thought processes and to discover the ways and means humans acquire recognition knowledge. After that, we can infer this knowledge to the computer to create more intelligent computational recognizers. This talk aims to share our findings with you related to the recognition of handwritten characters. It summarizes the results of several experiments we conducted in the past while attempting to understand the way humans write and recognize handwritten characters. Our investigations include handwriting education in elementary schools, handwriting models, stroke sequences, the legibility of different character shapes, left-handedness and right-handedness, the creation of databases for learning and testing, the derivation of the boundary between similar samples, and the pitfalls of current recognition algorithms and remedies. This talk concludes with highlights on the results of these studies and their applications to improve the reliability of computer recognition of handwritten characters.

Chinese fonts are recognized by a new method based on empirical mode decomposition. Five basic strokes have been selected to characterize the features of Chinese fonts. Based on them, stroke feature sequences of a given text block are calculated. Once decomposed by EMD, the first two intrinsic mode functions corresponding to each stroke feature sequence are used to calculate the stroke energy of all the five basic strokes. These energies are combined with the five averages of the residues to produce a ten-dimensional feature vector. Finally, the minimum distance classifier is used to recognize the fonts. Experiments show encouraging recognition rates.

The recognition of handwritten numerals is a challenging task in pattern recognition. It can be considered as one of the benchmarks in evaluating feature extraction methods and the performance of classifiers. In this paper, we propose a new method to improve the recognition accuracy of handwritten numerals by using hybrid feature extraction and random feature selection. First, we present seven feature extraction methods. A novel multi-class divergence criterion for large scale feature analysis is proposed and a random feature selection strategy is used to congregate three new hybrid feature sets. The new congregated features are complementary as they are formed from different original feature sets extracted by different means. Experiments conducted on MNIST database show that our proposed method can increase the recognition accuracy.

This paper describes a new approach to recognize touching numeral strings. Currently most methods for numeral string recognition require segmenting the string image into separate numerals. As a result, the recognition system heavily depends on the reliability of the segmentation module. This study explores the holistic strategy directly on the string images without segmentation. It builds the novel classifier by combining binary classifiers based on data-driven error correcting output coding (DECOC). The dimensions of input images are reduced using principal components analysis. Support vector machines are used as base learners. Experiments on NIST SD19 touching numeral pairs confirm that DECOC can achieve favorable performance compared with other multi-class holistic classifiers. The method provides the flexibility of controlling the computational complexity versus accuracy. We also discuss an implementation suitable for distributing computing by decomposing the ensemble into subtasks.

A fast iteration algorithm is proposed to approximate the reduced set vectors shared by each binary SVM solution for multi-class classification simultaneously. The iteration algorithm can be applied to the general kernel types such as k(/spl par/ x - x' /spl par//sup 2/) and k(x/sup T/x'). In addition, we present a fast block algorithm in the test phase to speed up the classification further. Experimental results have shown that the classification speeds on MNIST and Hanwang handwritten digit databases on P4 1.7 Ghz were about 16,000 and 10,895 patterns per second without sacrificing the classification accuracy of the original SVM system. The speed-up factor of 110 on MNIST database has been achieved.

Training a support vector machine on a data set of huge size with thousands of classes is a challenging problem. This paper proposes an efficient algorithm to solve this problem. The key idea is to introduce a parallel optimization step to quickly remove most of the nonsupport vectors, where block diagonal matrices are used to approximate the original kernel matrix so that the original problem can be split into hundreds of subproblems which can be solved more efficiently. In addition, some effective strategies such as kernel caching and efficient computation of kernel matrix are integrated to speed up the training process. Our analysis of the proposed algorithm shows that its time complexity grows linearly with the number of classes and size of the data set. In the experiments, many appealing properties of the proposed algorithm have been investigated and the results show that the proposed algorithm has a much better scaling capability than Libsvm, SVMlight, and SVMTorch. Moreover, the good generalization performances on several large databases have also been achieved.

This paper presents a novel approach for the verification of the word hypotheses generated by a large vocabulary, offline handwritten word recognition system. Given a word image, the recognition system produces a ranked list of the N-best recognition hypotheses consisting of text transcripts, segmentation boundaries of the word hypotheses into characters, and recognition scores. The verification consists of an estimation of the probability of each segment representing a known class of character. Then, character probabilities are combined to produce word confidence scores which are further integrated with the recognition scores produced by the recognition system. The N-best recognition hypothesis list is reranked based on such composite scores. In the end, rejection rules are invoked to either accept the best recognition hypothesis of such a list or to reject the input word image. The use of the verification approach has improved the word recognition rate as well as the reliability of the recognition system, while not causing significant delays in the recognition process. Our approach is described in detail and the experimental results on a large database of unconstrained handwritten words extracted from postal envelopes are presented.

Classification of sequential data using discriminative models such as support vector machines is very hard due to the variable length of this type of data. On the other hand, generative models such as HMMs have become the standard tool for representing sequential data due to their efficiency. This paper proposes a general generative-discriminative framework that uses HMMs to map the variable length sequential data into a fixed size P-dimensional vector (likelihood score) that can be easily classified using any discriminative model. The preliminary experiments of the framework on the MNIST database for handwritten digits have achieved a better recognition rate of 98.02% than that of standard HMMs (94.19%).

Previous research efforts on optical font recognition have mostly limited applications since they deal with only a few types of font attributes and estimate them from a line or block of text. This paper proposes a word-level optical font recognition system for printed Korean and English documents. At the word-level, it has the advantages of obtaining more detailed font attributes including the following: script (Korean and English), font style (regular, bold, italic, and underlined), typeface (Myung-jo and Gothic), point size (10, 12, 14 pts), and word length (2, 3, 4, 5 for Korean, and 4 to 10 for English). A hierarchical classifier and several typographical features have been devised for the system, and their effectiveness are proven by an experiment with a database of 100 sets of 264 font categories.

This paper investigates different strategies allowing integrationof contextual information during the feature extractionstage of a cursive handwriting HMM-based recognitionsystem. First we propose to use linear discriminant analysis(LDA) in order to integrate the class information duringfeature set building. Secondly several zoning strategies areused to integrate local contextual information. Finally, aweighting technique is proposed in association with zoningwith the aim of integrating handwriting style. Some experimentswere carried out and the results show the interest ofthe proposed strategies.

In this paper, we present a method of recognizinginclined, rotated characters. First we construct an eigensub-space for each category using the covariance matrixwhich is calculated from a sufficient number of rotatedcharacters. Next, we can obtain a locus by projectingtheir rotated characters onto the eigen sub-space andinterpolating between their projected points. An unknowncharacter is also projected onto the eigen sub-space ofeach category. Then, the verification is carried out bycalculating the distance between the projected point ofthe unknown character and the locus. In our experiment,we obtained quite good results for the CENTURY font of26 capital letters of the English alphabet (A, B, .... ,Z).This method has the added advantage of obtaining therecognition result (category) and angle of inclination atthe same time

In this paper a word recognition and verification scheme based on HMMs is presented. However, the main contribution of the current work lies in the validation of such a strategy. In order to perform this task, we carried out some experiments on word recognition using a legal amount database and then we compared the results reached with other study which makes use of the same database. The experiments demonstrate the efficiency of the strategy we developed for word recognition and verification.

... learning is proposed. It makes use of a multiobjective genetic algorithm where the minimization of the number of features and a validity index that measures the quality of clusters have been used to guide the search towards the more discriminant features and the best number of clusters. The proposed strategy is evaluated using two synthetic data sets and then it is applied to handwritten month word recognition. Comprehensive experiments demonstrate the feasibility and efficiency of the proposed methodology.

Feature selection for ensembles has shown to be an effective strategy for ensemble creation. In this paper we present an ensemble feature selection approach based on a hierarchical multi-objective genetic algorithm. The first level performs feature selection in order to generate a set of good classifiers while the second one combines them to provide a set of powerful ensembles. The proposed method is evaluated in the context of handwritten digit recognition, using three different feature sets and neural networks (MLP) as classifiers. Experiments conducted on NIST SD19 demonstrated the effectiveness of the proposed strategy.

We present a method for automatically selecting the best filter to treat poor quality printed documents using image quality assessment. We introduce five quality measures to obtain information about the quality of the images, and morphological filters to improve their quality. A training set of 370 images was used to develop the system. Experimental results on the test set show a significant improvement in the recognition rate from 73.24% using no filter at all to 93.09% after applying a filter that was automatically selected.

This paper summarizes the research activities of the pastdecade on the recognition of handwritten scripts used inChina, Japan, and Korea. It presents the recognitionmethodologies, features explored, databases used, andclassification schemes investigated. In addition, it includes adescription of the performance of numerous recognitionsystems found in both academic and industrial researchlaboratories. Recent achievements and applications are alsopresented. A list of relevant references is attached togetherwith our remarks on this subject.

This paper describes a system being developed to recognizedate information handwritten on Canadian bankcheques. A segmentation based strategy is adopted in thissystem. In order to achieve high performances in terms ofefficiency and reliability, a knowledge-based module is proposedfor the date segmentation and a cursive month wordrecognition module is implemented based on a combinationof classifiers. The interaction between the segmentation andrecognition stages is properly established by using multi-hypothesesgeneration and evaluation modules. As a result,promising performance is obtained on a test set from a real-lifestandard cheque database.

A fast support vector machine (SVM) training algorithm is proposed under SVM's decomposition framework by effectively integrating kernel caching, digest and shrinking policies and stopping conditions. Kernel caching plays a key role in reducing the number of kernel evaluations by maximal reusage of cached kernel elements. Extensive experiments have been conducted on a large handwritten digit database MNIST to show that the proposed algorithm is much faster than Keerthi et al.'s improved SMO, about nine times. Combined with principal component analysis, the total training for ten one-against-the-rest classifiers on MNIST took less than an hour. Moreover, the proposed fast algorithm speeds up SVM training without sacrificing the generalization performance. The 0.6% error rate on MNIST test set has been achieved. The promising scalability of the proposed scheme paves a new way to solve more large-scale learning problems in other domains such as data mining.

In this paper a methodology for feature selection for the handwritten digit string recognition is proposed. Its novelty lies in the use of a multiobjective genetic algorithm where sensitivity analysis and neural network are employed to allow the use of a representative database to evaluate fitness and the use of a validation database to identify the subsets of selected features that provide a good generalization. Some advantages of this approach include the ability to accommodate multiple criteria such as number of features and accuracy of the classifier, as well as the capacity to deal with huge databases in order to adequately represent the pattern recognition problem. Comprehensive experiments on the NIST SD19 demonstrate the feasibility of the proposed methodology.

This paper is an improvement on the characterization of edges. Using a novel wavelet function, it is proven that the maximum moduli of the wavelet transform (MMWT) of a curve produces two new symmetrical curves on both sides of the original with the same direction. The distance between the two curves is shown to be independent of the width d of the original curve if the scale s of the wavelet transform satisfies s/spl ges/d. This property provides a novel method of obtaining the skeletons of the curves in an image.

This paper describes a fast two-level Viterbi search algorithm for recognizing handwritten words as a sequence of characters concatenated according to a lexicon. The algorithm is based on hidden Markov model (HMM) representations of characters and it breaks up the computation of word likelihood scores into two levels: state level and character level. This enables the reuse of likelihood scores of characters to decode all words in the lexicon, avoiding repeated computation of state sequences. Experimental results with an 85,000-word vocabulary indicate that the computational cost of an off-line handwritten word recognition system may be reduced by more than a factor of 20 while not introducing search errors.

A modular system to recognize handwritten numerical strings is proposed. It uses a segmentation-based recognition approach and a recognition and verification strategy. The approach combines the outputs from different levels such as segmentation, recognition, and postprocessing in a probabilistic model. A new verification scheme which contains two verifiers to deal with the problems of oversegmentation and undersegmentation is presented. A new feature set is also introduced to feed the oversegmentation verifier. A postprocessor based on a deterministic automaton is used and the global decision module makes an accept/reject decision. Finally, experimental results on two databases are presented: numerical amounts on Brazilian bank checks and NIST SD19. The latter aims at validating the concept of modular system and showing the robustness of the system using a well-known database.

In this paper, we present a document analysis system which is expected to extract regions of interest in greyscale document images. Collected areas are then clustered in text zones and non-text areas using geometric and texture features. The system works in two steps. Regions of interest are retrieved via cumulative gradient considerations. In classification module, we introduced some entropic heuristic. Experiments are done on the MediaTeam Document Database to show the relevance of this criteria.

A new direction in machine learning area has emerged from Vapnik's theory in support vectors machine (SVM) and its applications on pattern recognition. In this paper we propose a new SVM kernel family, called KMOD (kernel with moderate decreasing) with distinctive properties that allow better discrimination in the feature space. The experiments that we carry out show its effectiveness on synthetic and large-scale data. We found KMOD performs better than RBF and exponential RBF kernels on the two-spiral problem. In addition, a digit recognition task was processed using the proposed kernel. The results show, at least, comparable performances to state of the art kernels.

The authors propose a handwritten numeral string recognition method composed of two HMM-based stages. The first stage uses an implicit segmentation strategy based on string contextual information to provide multiple segmentation-recognition paths. These paths are verified and re-ranked by using a verification stage based on a digit classifier. It allows the use of two sets of features and numeral models: one taking into account both segmentation and recognition aspects in an implicit segmentation based strategy, and another considering just recognition aspects of isolated digits. The two system stages are shown to be complementary in the sense that the verification stage is shown to be a promising idea to deal with the loss in terms of recognition performance brought about by the necessary tradeoff between segmentation and recognition carried out in the first system stage.

This paper proposes a general local learning framework to effectively alleviate the complexities of classifier design by means of "divide and conquer" principle and ensemble method. The learning framework consists of quantization layer and ensemble layer. After GLVQ and MLP are applied to the framework, the proposed method is tested on MNIST handwritten digit database. The obtained performance is very promising, an error rate with 0.99%, which is comparable to that of LeNet5, one of the best classifiers on this database. Further, in contrast to LeNet5, our method is especially suitable for a large-scale real-world classification problem.

During the development of a hidden Markov model based handwriting recognition system, the testing phase takes a non-negligible amount of computation time. This is especially true for real application where the lexicon size is large. In order to shorten the development process, we propose an indicator of the system discrimination power. This indicator is calculated during training and its final value is obtained at the end of the training phase, without more calculation. Its definition consists of a modification of the observation probability of the validation corpus by the trained system. Some experiments were carried out and the results show clearly the correlation between this indicator and recognition rates.

A realignment algorithm for irregular character strings on color documents is proposed. Color documents often contain poorly aligned texts such as inclined or curved texts sometimes with distortion. In order to recognize them, we classify these texts into five types. After determining the type, we realign all the characters in a text horizontally, then test them with an ordinary character recognition method. Lastly, we show some experimental results for texts extracted from real color documents and discuss some causes of misrecognition.

This paper presents recognition of unconstrained handwritten numeral strings using a decision value generator. The numeral string recognition system is composed of three modules: pre-segmentation, segmentation and recognition. The pre-segmentation module classifies a numeral string into sub-images, such as isolated digits, touching digits or broken digits, based on the confidence value of decision value generator. The segmentation module splits the touching digits using the reliability value of decision value generator. Both segmentation-based and segmentation free methods are used in classification and segmentation. To evaluate the proposed method, experiments were conducted using the handwritten numeral strings of NIST SD19 and a higher recognition performance than previous works was obtained.

A sophisticated methodology of legal amount recognition based on the word segmentation hypotheses is introduced for automatic bank check processing. Word segmentation hypotheses are derived according to the grapheme level segmentation results of the legal amount. Novel hybrid schemes of HMM-MLP classifiers are also introduced for producing the ordered legal word recognition results with reliable decision values. These values can be used for obtaining an optimal word segmentation path of over-segmentation hypotheses as well as an efficient rejection criterion of word recognition result. Simulation was performed with CENPARMI bank check database and shows quite encouraging results.

We propose a word segmentation method for handwritten Korean text lines. It uses gap information to separate a text line into word units, where the gap is defined as a white-run obtained after a vertical projection of the line image. Each gap is classified into a between-word gap or a within-word gap using a clustering technique. We take up three gap metrics - the bounding box (BB), run-length/Euclidean (RLE) and convex hull (CH) distances - which are known to have superior performance in Roman-style word segmentation, and three clustering techniques - the average linkage method, the modified MAX method and sequential clustering. An experiment with 498 text-line images extracted from live mail pieces has shown that the best performance is obtained by the sequential clustering technique using all three gap metrics.

Many offline handwritten word recognition systems have been proposed since the early nineties. Most systems reported high recognition rates, however, they overlooked a very important factor in the process: speed factor. The authors explore the potential for speeding up an offline handwritten word recognition system via concurrency. The goal of the system is to achieve both full accuracy and high speed when taking into account large vocabularies. This was accomplished by integrating the recognition process with multiprocessing and distributed computing concepts. Experimental results showed that the multiprocessing environment is very promising in enhancing a sequential offline handwritten word recognition system performance.

This paper describes an off-line system under development to process unconstrained handwritten dates on Brazilian bank cheques in an omni-writer context. We show here some improvements on our previous work on isolated month word recognition using hidden Markov models (HMM). After preprocessing, a word image is explicitly segmented into characters or pseudo-characters and represented by two feature sequences of equal length, which are combined using HMM. The word models are generated from the concatenation of appropriate character models. In addition to the small date database, we also make use of the legal amount database to increase the frequency of characters in the training and the validation sets. Although this study deals with a limited lexicon, the many similarities among the word classes can affect the performance of the recognition. Experiments show an increase in the average recognition rate from 84% to 91%. Finally, we present our perspectives of future work.

A class-modular classifier can be characterized by two prominent features: low classifier complexity and independence of classes. While conventional character recognition systems adopting the class modularity are faithful to the first feature, they do not investigate the second one. Since a class can be handled independently of the other classes, the class-specific feature set and classifier architecture can be optimally designed for a specific class Here we propose a general framework for the class modularity that exploits fully both features and present four types of class-modular architecture. The neural network classifier is used for testing the framework A simultaneous selection of the feature set and network architecture is performed by the genetic algorithm. The effectiveness of the class-specific features and classifier architectures is confirmed by experimental results on the recognition of handwritten numerals.

The paper presents a modular system to recognize numerical amounts on Brazilian bank cheques. The system uses a segmentation-based recognition approach and the recognition function is based on a recognition and verification strategy. Our approach consists of combining the outputs from different levels such as segmentation, recognition and post-processing in a probabilistic model. A new feature set is introduced to the verifier module in order to detect segmentation effects such as over-segmentation and under-segmentation. Finally, we present experimental results on two databases: numerical amounts and NIST SD19. The latter aims at validating the concept of modular system and showing the robustness of the system over a well-known database.

We introduce a technique based on diagonal white runs and vertical edges, that divides a document image into columns and blocks which are subsequently classified as text or graphics. A diagonal white run (drun) is a set of adjacent white pixels that are diagonally connected, and a vertical edge consists of the white area between two consecutive druns. This technique was designed as a layout-independent approach. Testing the proposed approach on document images with 14 different types and layouts, written in different languages, shows comparative and promising results.

Segmenting handwritten date fields on bank cheque images into three subimages corresponding to the day, month and year is the first and critical step of our date recognition system. The paper describes a knowledge-based segmentation system, which introduces different kinds of knowledge at different segmentation stages to improve the performance. The knowledge includes information on the writing style, syntactic and semantic constraints, etc. Results have shown that the system is very effective compared with a previous structural feature based method.

Although promising results on the combination of character recognizers have been reported recently, the combination strategies can not be readily applied to the recognition of character strings due to m-n correspondence problems caused by segmentation errors. In this paper, we propose a new paradigm of combining multiple string recognizers and contribute a generic framework for off-line combination. We designed and implemented a graph based off-line combination system, StrCombo, which has achieved a substantial improvement over any one of the individual recognizers in a real-life application. This open combination system provides the possibility of further improving the performance of string recognizers when new recognizers and combination rules are available.

The proposed feedback-based approach is implemented in two steps. In the first step, segmentation is done according to the structural features between the connected components in the legal amounts. In the second step, a feedback process is introduced to re-segment the parts that could not be identified in the first step. Then a multiple neural network classifier is used to verify the re-segmentation result. The confidence value produced by the classifier is used to determine the best segmentation points. This approach is tested on a CENPARMI database and the result indicates that the correct segmentation rate increased by 13.4% from the previous approach.

Global gray-level thresholding techniques such as Otsu's method, and local gray-level thresholding techniques such as edge-based segmentation or the adaptive thresholding method are powerful in extracting character objects from simple or slowly varying backgrounds. However, they are found to be insufficient when the backgrounds include sharply varying contours or fonts in different sizes. A stroke-model is proposed to depict the local features of character objects as double-edges in a predefined size. This model enables us to detect thin connected components selectively, while ignoring relatively large backgrounds that appear complex. Meanwhile, since the stroke width restriction is fully factored in, the proposed technique can be used to extract characters in predefined font sizes. To process large volumes of documents efficiently, a hybrid method is proposed for character extraction from various backgrounds. Using the measurement of class separability to differentiate images with simple backgrounds from those with complex backgrounds, the hybrid method can process documents with different backgrounds by applying the appropriate methods. Experiments on extracting handwriting from a check image, as well as machine-printed characters from scene images demonstrate the effectiveness of the proposed model.

A methodology of combining HMM (hidden Markov model) and MLP (multilayer perceptron) for cursive word recognition is presented in this paper. We have designed an explicit segmentation based HMM, and combined it with an implicit segmentation based MLP using weighting coefficients. The main idea of this methodology is that more distinct classifiers can better complement each other. We also introduced a new probability measure for the hybrid classifier as well as conventional combining schemes. Experiments were conducted with month word and legal word databases of CENPARMI and improved performances of 87.3% for 21 month word classes and 92.2% for 32 legal word classes have been achieved.

Describes a scheme for recognizing unconstrained handwritten numeral strings by a composite segmentation method which combines both recognition-free and recognition-based segmentation methods. A digit group detector has been designed to separate touching digits from isolated digits by the recognition-free segmentation method. Subsequently touching digits are segmented by prioritized segmentation points accomplished by analyzing the ligature and touching types. Four special kinds of candidate segmentation points and six touching types are defined to obtain more stable segmentation points. To evaluate the proposed method, we have experimented with 1,500 numeral strings of the NIST SD19 database and obtained a recognition rate of 91.8%.

Clustering combination uses more than one clustering method with identical pattern features to improve the clustering performance. In general clustering is an optimization procedure based on a specific clustering criterion, so clustering combination can be regarded as a technique that constructs and processes multiple clustering criteria rather than a single criterion. We propose two methods of combining objective function clustering and graph theory clustering. One incorporates multiple criteria into an objective function according to their importance, and solves this problem with constrained nonlinear optimization programming. The other method consists of two sequential procedures: (a) a traditional objective function clustering for generating the initial result, and (b) an autoassociative additive system based on graph theory clustering for modifying the initial result.

The last frontiers of handwriting recognition are considered to have started in the last decade of the second millennium. The paper summarizes (a) the nature of the problem of handwriting recognition, (b) the state of the art of handwriting recognition at the turn of the new millennium, and (c) the results of CENPARMI researchers in automatic recognition of handwritten digits, touching numerals, cursive scripts, and dates formed by a mixture of the former 3 categories. Wherever possible, comparable results have been tabulated according to techniques used, databases, and performance. Aspects related to human generation and perception of handwriting are discussed. The extraction and usage of human knowledge, and their incorporation into handwriting recognition systems are presented. Challenges, aims, trends, efforts and possible rewards, and suggestions for future investigations are also included.

The generically used expert scheduling system (GUESS) intelligent scheduling toolkit has been built and applied in various scheduling domains. Previously, GUESS included three major scheduling approaches: a heuristic‐based (suggestion tabulator) approach, a hill‐climbing algorithm, and a genetic algorithm approach. GUESS has now been expanded to include a neural network scheduling approach. This paper discusses the development, implementation and testing results of the neural network scheduling method within GUESS.

This paper describes a novel method for edge feature detection of document images based on wavelet decomposition and reconstruction. By applying the wavelet decomposition technique, a document image becomes a wavelet representation, i.e. the image is decomposed into a set of wavelet approximation coefficients and wavelet detail coefficients. Discarding wavelet approximation, the edge extraction is implemented by means of the wavelet reconstruction technique. In consideration of the mutual frequency, overlapping will occur between wavelet approximation and wavelet details, a multiresolution-edge extraction with respect to an iterative reconstruction procedure is developed to ameliorate the quality of the reconstructed edges in this case. A novel combination of this multiresolution-edge results in clear final edges of the document images. This multi-resolution reconstruction procedure follows a coarser-to-finer searching strategy. The edge feature extraction is accompanied by an energy distribution estimation from which the levels of wavelet decomposition are adaptively controlled. Compared with the scheme of wavelet transform, our method does not incur any redundant operation. Therefore, the computational time and the memory requirement are less than those in wavelet transform.

Two strategies can be considered in handwriting recognition: phrase or word approaches. In this paper, we demonstrate the superiority of the phrase-based strategy, especially in city name recognition. The performance of an HMM-based off-line system using an analytic approach with explicit segmentation is evaluated on two databases: (i) city names in full, and (ii) city names in single words. A difference in performance is observed, principally caused by the dissimilarity of word lengths between the two databases. After generating other data sets and lexicons, experiments were performed yielding results which lead us to conclude that word length in the data set, as well as in lexicons, significantly influences recognition performance, and also that it is preferable to perform city name recognition based on the phrase approach rather than by word recognition.

We propose a new method for the recognition of unconstrained handwritten words consisting of Korean and numeric characters. To overcome the difficulty of separating touching characters, we adopt an over-segmentation technique and we find the optimal segment combination using a lexicon-driven word scoring technique and a nearest neighbor classifier. The optimal combination gives the final segmentation positions for individual characters with the best matching word in the lexicon. The proposed system has yielded an accuracy of 90.64% for 908 word images on live mail pieces.

The paper introduces a method of finding the neighborhood of the optimal number of hidden neurons for an error backpropagation neural network with a single hidden layer. It is based on a study of the curvature of the error function, during the training phase of the network. The method assures convergence and bypasses local minimas. Experimental results show the uniqueness of the method's solution regardless of the initial values of the network's parameters. Two neural networks were built, one for recognizing unconstrained handwritten English numerals and the other for Arabic numerals. Recognition results and comparison with other methods are also presented.

This paper proposes a model-based character extraction method. We model a pixel belonging to a character as a double-edge, whose range is defined by the stroke width, and whose intensity is proportional to the local contrast. By extracting such double-edge feature at a predefined stroke width, sharply changing as well as slowly varying backgrounds can be eliminated. A set of morphological operators is designed to extract the double-edge feature at each pixel of the raw image, and the characters are extracted by thresholding these features. A goal directed evaluation of extraction of the courtesy amount from bank cheques reveals the advantages of the proposed method over other existing methods. Visual inspection of legal amount extraction shows further promise in extracting characters from complex backgrounds.

Describes a hidden Markov model-based approach designed to recognize off-line unconstrained handwritten words for large vocabularies. After preprocessing, a word image is segmented into letters or pseudoletters and represented by two feature sequences of equal length, each consisting of an alternating sequence of shape-symbols and segmentation-symbols, which are both explicitly modeled. The word model is made up of the concatenation of appropriate letter models consisting of elementary HMMs and an HMM-based interpolation technique is used to optimally combine the two feature sets. Two rejection mechanisms are considered depending on whether or not the word image is guaranteed to belong to the lexicon. Experiments carried out on real-life data show that the proposed approach can be successfully used for handwritten word recognition.

This paper describes techniques for stroke extraction used in the recognition of handwritten Chinese characters. A new set of feature points is proposed for the analysis of skeleton images. Based on a geometrical graph, a novel criterion is proposed for the identification of fork points in a skeleton image which correspond to joint points in the original character image. Experimental results indicate that the proposed method correctly determines the fork points, and is effective in unifying the joint points.

In this paper, we propose a new approach to combine multiple features in handwriting recognition based on two ideas: feature selection-based combination and class dependent features. A nonparametric method is used for feature evaluation, and the first part of this paper is devoted to the evaluation of features in terms of their class separation and recognition capabilities. In the second part, multiple feature vectors are combined to produce a new feature vector. Based on the fact that a feature has different discriminating powers for different classes, a new scheme of selecting and combining class-dependent features is proposed. In this scheme, a class is considered to have its own optimal feature vector for discriminating itself from the other classes. Using an architecture of modular neural networks as the classifier, a series of experiments were conducted on unconstrained handwritten numerals. The results indicate that the selected features are effective in separating pattern classes and the new feature vector derived from a combination of two types of such features further improves the recognition rate.

Describes the latest developments to enhance the performance of our HMM-based handwritten word recognition system. These methods only deal with the recognition phase and involve the improvement of the HMM architecture as well as the optimization of the training phase. Experiments carried out on real data show that the proposed approaches lead to significant improvements in the accuracy of the system.

Describes the mixed HMM-KNN word recognition module of a bank cheque processing system developed at CENPARMI. It uses a combination of 2 segmentation free word recognition schemes. The first scheme uses a set of global features associated to a modified K nearest neighbour classifier; while the second one uses a set of directional contour features as input to an HMM. The system has been designed to be modular and independent of specific languages as in Canada one has to deal with at least 2 languages, namely English and French. It can be easily adapted to read other European languages based on the Roman alphabet. The system is continuously tested on data from the local phone company, and we report here the results on a database of approximately 4,500 cheques.

We analyze the class separation of the features in handwriting recognition. Behaviors of measurement tools are studied with a partial and full classifications. A new scheme of selecting and combining class-dependent features is proposed. In this scheme, a class is considered to have its own optimal feature vector for discriminating itself from the other classes. Using an architecture of modular neural networks as the classifier, a series of experiments have been conducted on totally unconstrained handwritten numerals. The results indicate that the selected features are effective in separating pattern classes and the new feature vector derived from a combination of two types of such features further improves the recognition rate.

Automatic processing of international documents presents a number of challenging problems because Optical Character Recognition (OCR) techniques are not available for all languages and all script classes. Document images must be categorized according to their script type first, in our case Roman, Ideographic, or Arabic. We present a set of statistical methods that first detect and correct the skew of a document image. Next, the page is segmented into text and graphical components. The textual components are then segmented into paragraphs and lines; and finally we classify the script type into one of three categories. The system predicts the correct script category in 91% of cases when tested on real-life documents of varying kinds, diverse formats and qualities from many sources.

This paper proposes a novel method which enables a Chinese character recognition system to obtain reliable recognition. In this method, two thresholds, i.e. class region thresholdRk and disambiguity thresholdAk, are used by each Chinese character k when the classifier is designed based on the nearest neighbor rule, where Rk defines the pattern distribution region of character k, and Ak prevents the samples not belonging to character k from being ambiguously recognized as character k. A novel algorithm to derive the appropriate thresholds Ak and Rk is developed so that a better recognition reliability can be obtained through iterative learning. Experiments performed on the ITRI printed Chinese character database have achieved highly reliable recognition performance (such as 0.999 reliability with a 95.14% recognition rate), which shows the feasibility and effectiveness of the proposed method.

Two types of techniques are usually adopted in language differentiation: token matching and statistical analysis. In this paper we present a method which uses a combined analysis of several discriminating statistical features, for the differentiation between European and oriental language scripts. When applied to more than 23 languages, it has proved to be effective in differentiating between documents printed in these different scripts.

This paper includes a description of 3 affiliated oriental languages: Chinese, Japanese, and Korean. It includes a description of the origins of these 3 languages and the inter-relationship among them. Drawn from the viewpoints of several experienced researchers in the field of OCR (Optical Character Recognition) and computational linguistics, it attempts to bring out the intriguing aspects of these 3 ideographic languages, including the formation and composition of pictograms, special features, learning, understanding, contextual information, and recognition of characters and words, and their relations to poetic expressions and pattern recognition techniques. Numerous references are given and comments on future trends are also presented.

We introduce the new concept of "veinerization", which produces a graph that contains all the "topological" information needed to derive a wide variety of skeletons. Theoretically, the main contribution is to provide a homogeneous framework for integration of the major concepts described in other related works on digital skeletonization. In practice, the new aspect of this approach is to provide the user with different criteria for selecting the most suitable skeleton for a given application, e.g., the user can select a suitable threshold for obtaining the desirable balance between " having a skeleton without noisy prunes" and "having a skeleton that reflects the initial shape". This algorithm has been tested on numerous kinds of patterns, including pathological ones like fractal sets well-known for the complexity of their shapes.

In this correspondence, we present a general recursive approach for image segmentation by extending Otsu's (1978) method. The new approach has been implemented in the scope of document images, specifically real-life bank checks. This approach segments the brightest homogeneous object from a given image at each recursion, leaving only the darkest homogeneous object after the last recursion. The major steps of the new technique and the experimental results that illustrate the importance and the usefulness of the new approach for the specified class of document images of bank checks will be presented.

Two patterns are matched by putting one on top of the other and iteratively moving their individual parts until most of their corresponding parts are aligned. An energy function and a neighborhood of influence are defined for each iteration. Initially, a large neighborhood is used such that the movements result in global features being coarsely aligned. The neighborhood size is gradually reduced in successive iterations so that finer and finer details are aligned. Encouraging results have been obtained when applied to match complex Chinese characters. It has been observed that computation increases with the square of the number of moving parts which is quite favorable compared with other algorithms. The method was applied to the recognition of handwritten Chinese characters. After performing the iterative matching, a set of similarity measures are used to measure the similarity in topological features between the input and template characters. An overall recognition rate of 96.1% is achieved.

A new piecewise linear modulation model of handwriting is proposed. In this model, the velocity of handwriting trajectory is modeled as the impulse response of a time varying second order system. For mathematical tractability, the entire trajectory is segmented into several non-overlapping frames, while the natural frequency and the damping factor of the system are assumed to vary linearly with time in each frame and are continuous along the entire trajectory. In other words, handwriting is regarded as an oscillation modulated by a continuous and piecewise linear signal. The parameters of this model are estimated by Powell's optimization algorithm which does not require the computation of the first order derivative. The number and lengths of the frames are decided by a modified binary search algorithm along with the estimation of parameters. This model has achieved very high data compression rate as well as accurate reproduction of real handwriting.

Presents a model for reading cursive scripts which has an architecture inspired by a reading model and which is based on perceptual concepts. We limit the scope of our study to the off-line recognition of isolated cursive words. First of all, we justify why we chose McClelland & Rumelhart's (1981) reading model as the inspiration for our system. A brief resume/spl acute/ of the method's behavior is presented and the main originalities of our model are underlined. After this, we focus on the new updates added to the original system: a new baseline extraction module, a new feature extraction module and a new generation, validation and hypothesis insertion process. After implementation of our method, new results have been obtained on real images from a training set of 184 images, and a testing set of 100 images, and are discussed. We are concentrating now on validating the model using a larger database.

Two types of techniques are usually adopted in language differentiation: token matching and statistical analysis. In this paper we present a method which uses a combined analysis of several discriminating statistical features for the differentiation between European and oriental language scripts. When applied to more than 23 languages, it has proved to be effective in classifying documents printed in these different scripts.

We describe a hidden Markov model (HMM) based word recognition engine being developed to be integrated with the CENPARMI bank cheque processing system. The various modules are described in detail, and preliminary results are compared with our previous global feature recognition scheme. The engine is tested on words from a database of over 4,500 cheques of 1,400 writers.

Presents a robust thinning-based method for the segmentation of strokes from handwritten Chinese characters. A new set of feature points is proposed for the analysis of skeleton images. A geometrical graph-based approach is developed for the analysis of strokes. A novel criterion is proposed for the identification of the fork points in a skeleton image which correspond to the same joint points in the original character image. Experimental results show that the proposed method is effective.

The authors have extended existing methods to identify the language of an on-line document after the characters have been coded using 10 character classes based on visual characteristics. In particular, they exploit word bigrams and trigrams in both a linear combination of score values and an expert systems approach. Knowledge about each language as acquired from a large number of on-line texts. Using a small set of rules, the expert system outperforms the linear combination in accuracy and shows more stability when parameter settings are varied.

The performance of a character recognition system depends heavily on what features are being used. Though many kinds of features have been developed and their test performances on a standard database have been reported, there is still room to improve the recognition rate by developing an improved feature. The authors propose a new feature based on DDD (directional distance distribution) information. This new concept regards the input pattern array as being circular. It also contains very rich information by encoding in one representation both the white/black distribution and the directional distance distribution. A test performed on the CENPARMI handwritten numeral database showed a promising result of 97.3% recognition with a neural network classifier using the DDD feature.

This paper describes a Chinese cheque processing system currently under development at the Centre for Pattern Recognition and Machine Intelligence (CENPARMI). The information on Chinese bank cheques is not the same as that on alphanumeric bank cheques. The legal amount in a Chinese bank cheque is the Chinese character text associated with each currency unit. This paper discusses a technique using each currency unit as a key word to locate/extract the legal amount in bank cheques. In the analysis and recognition process, the system tries to locate the smallest currency units in the image and identifies it first. Then, the system tries to locate the image strings associated with each currency unit. Each image string is separated and recognized. Next, a set of rules and context are applied to recognize the characters. In order to choose the correct one, the recognized character string is accepted only if it satisfies all the conditions governed by rules.

The paper describes a high performance offline system for recognizing hand printed numerals. An innovative verification module is applied which drastically improves the recognition rate. The approaches used in the modules are described. The importance of the verification module is analysed in detail. A practical automatic form reading system TOCR V1.0 was developed based on the algorithms. The system was put into practical use in several provinces of China for statistical analysis of Revenue China. Test results are given based on: 1) data collected when the system was used in China, as well as 2) the CENPARMI database.

A novel approach to extract data from check images is proposed based on the determination of baselines of checks, a priori information about the positions of data on checks, and a layout-driven item extraction method. Several techniques and algorithms have been developed in this approach including check image preprocessing, the extraction and identification of baselines, the extraction of the strokes of handwritten legal amounts, courtesy amounts and date, and the separation of strokes connected to baselines. A complete working system has been developed. The results of both testing experiments and on-line applications show that this approach is effective and the proposed techniques and algorithms perform well.

In several applications where binary contours are used to represent and classify patterns, smoothing must be performed to attenuate noise and quantization error. This is often implemented with local weighted averaging of contour point coordinates, because of the simplicity, low-cost and effectiveness of such methods. Invoking the "optimality" of the Gaussian filter, many authors will use Gaussian-derived weights. But generally these filters are not optimal, and there has been little theoretical investigation of local weighted averaging methods per se. This paper focuses on the direct derivation of optimal local weighted averaging methods tailored towards specific computational goals such as the accurate estimation of contour point positions, tangent slopes, or deviation angles. A new and simple digitization noise model is proposed to derive the best set of weights for different window sizes, for each computational task. Estimates of the fraction of the noise actually removed by these optimum weights are also obtained. Finally, the applicability of these findings for arbitrary curvature is verified, by numerically investigating equivalent problems for digital circles of various radii.

One of the key technologies related to knowledge and data engineering is the acquisition of knowledge and data in the development and utilization of information system and the strategies to capture new knowledge and data. Actually, millions of documents, including technical reports, government files, newspapers, books, magazines, letters, bank checks, etc., have to be processed every day, and knowledge has to be acquired from them. This paper presents a new approach to document analysis for automatic knowledge acquisition. The traditional approaches have two major disadvantages: (1) They are not effective for processing documents with high geometrical complexity. Specially, the top-down approach can process only the simple documents which have specific format or contain some a priori information. (2) The top-down approach needs to split large components into small ones iteratively, while the bottom-up approach needs to merge small components into large ones iteratively. They are time consuming. This new approach is based on modified fractal signature. It can overcome the above weaknesses.

It has been demonstrated that combining the decisions of several classifiers can lead to better recognition results. The combination can be implemented using a variety of strategies, among which majority vote is by far the simplest, and yet it has been found to be just as effective as more complicated schemes in improving the recognition results. This paper examines the mode of operation of the majority vote method in order to gain a deeper understanding of how and why it works, so that a more solid basis can be provided for its future applications to different data and/or domains. In the course of our research, we have analyzed this method from its foundations and obtained many new and original results regarding its behavior. Particular attention has been directed toward the changes in the correct and error rates when classifiers are added, and conditions are derived under which their addition/elimination would be valid for the specific objectives of the application. At the same time, our theoretical findings are compared against experimental results, and these results do reflect the trends predicted by the theoretical considerations.

A new method of optimizing prototypes for a nearest neighbor classifier is proposed based on a four-layer network architecture. A new error function is defined for updating prototypes. The physical meaning of the updating rule and the relationship between the proposed method and LVQ2 are also presented. The main characteristic of the present method is consistent criteria for updating prototypes and for using the trained prototypes to build a nearest neighbor classifier. Experimental results indicate that the present method is effective compared with LVQ2.

A novel approach is proposed for the extraction of legal and courtesy amounts and date from cheque images based on the structural description of cheques. A method for the representation of cheques is presented. Several image processing techniques and algorithms have been developed in this approach. Experimental results show that the approach is effective and the proposed techniques and algorithms perform well.

This paper describes a generic document segmentation and geometric relation labeling method with applications to document analysis. Unlike the previous document segmentation methods where text spacing, border lines, and/or a priori layout models based template processing are performed, the present method begins with a hierarchy of partitioned image layers where inhomogeneous higher-level regions are recursively positioned into lower-level rectangular subregions and at the same time lower-level smaller homogeneous regions are merged into larger homogeneous regions. The present method differs from the traditional split-and-merge segmentation method in that it orthogonally splits regions using thresholds adaptively computed from projection profiles.

In most document analysis and recognition systems, straight lines are considered as one of the basic elements that should be located and eliminated to simplify the process of document analysis and recognition. The superposition or the intersection of different objects of interest found in the same area makes the process of detecting and extracting these line segments a non-trivial task to pursue, especially, if the method should preserve the valuable objects of interest that intersect with these lines. In this paper, we present a new and effective approach that detects the existence of line segments and eliminates them with the challenge of preserving the valuable information that intersects these line segments. The new approach makes use of the well-known morphological processing technique of the closing operation, that uses a fixed structuring element, towards the use of a dynamic structuring element. The purpose of the new dynamic structuring element is to detect and preserve the valuable objects intersecting the line segments that should be eliminated regardless of the different orientations with which the objects intersect these line segments.

In this paper, we present a novel approach to optical character recognition that utilizes ring-projection-wavelet-fractal-signatures. In particular, the proposed approach reduces the dimensionality of a two-dimensional pattern by way of a ring-projection method, and thereafter, performs Daubechies' wavelet transform on the derived one-dimensional pattern to generate a set of wavelet sub-patterns, namely, curves that are non-self intersecting. Further from the resulting non-self intersecting curves, the divider dimensions are readily computed. These divider dimensions constitute a new characteristic vector for the original two-dimensional pattern, defined over the curves' fractal dimensions.

A new algebraic feature extraction method for image recognition is presented. The optimal transform of image matrices is proposed to extract the features from images. The Frobenius norm of matrices is first introduced as a measure of the distance between two matrices. Based on this, the within-class and between-class distances of image samples are defined. The ratio of the between-class and within-class distances of the transformed image sample set is taken as the criterion function J(T). The optimal transform matrix T is calculated by maximizing J(T) under some constraints. Experiments have been conducted to recognize both human face and handwritten character images. These results indicate that the algebraic features extracted by the present method possess a very strong discriminant power. An important conclusion about the present method is that the traditional linear discriminant method can be considered as a special case of the present feature method when image samples have only one column of vectors.

We present a formal model for processing gray-scale images of business forms such as bank cheques. The formal model is based on a new hybrid-based approach namely the base lines. In fact, to segment handwritten and hand-printed data from bank cheques, knowledge rules and base lines will have important roles to segment and extract the information from bank cheques. The architectural design as well as the major components of the system is discussed in full detail. Moreover, the significant use of the morphological followed by the topological processing on gray-scale images is used as a major aspect to restore the lost information after the elimination of the background and the base lines from the gray-scale cheques.

This paper presents a new perception based model for reading cursive script. We describe the organization of our pseudo-neuronal system and show the role of activation mechanism in perceiving and reading cursive script. We have introduced into our model some characteristics specific to cursive script. First, we use more appropriate features such as ascenders and descenders. Second, we deal with the ambiguity of letter location by introducing the concept of the fuzzy position. The location as well as the missing letters are deduced from the context (i.e. the word-letter lexicon). After implementation of our method, preliminary qualitative results have been obtained and are discussed. We are concentrating now on further formalizing and generalizing the proposed model on a larger data base.

A method for recognizing unconstrained handwritten words belonging to a small static lexicon is proposed. Our computational theory is based on a psychological model of the reading process of a fast reader. The method we propose is global in its nature and avoid the difficult segmentation stage of common word recognition techniques. Our computational theory has been applied to the processing of handwritten bank cheques, whose problem domain is that of unconstrained handwriting, unlimited writers in a small static lexicon. Current results seem comparable to those published in the literature and support our computational theory.

We propose a generic two-stage multi-network classification scheme and a realization of this generic scheme: a two-stage multi-network OCR system. The generic two-stage multi-network classification scheme decomposes the estimation of a posteriori probabilities into two coarse-to-fine stages. This generic classification scheme is especially suitable for the classification tasks which involve a large number of categories. The two-stage multi-network OCR system consists of a bank of specialized networks, each of which is designed to recognize a subset of whole character set. A soft pre-classifier and a network selector are employed in the two-stage multi-network OCR system for selectively invoking necessary specialized network. The network selector makes decisions based on both the prior case information and the outputs of the pre-classifier. Compared with the system which uses either a single network or one-stage multiple networks, the two-stage multi-network OCR system offers advantages in recognition accuracy, confidence measure, speed, and flexibility.

This paper presents a new approach to document analysis. The proposed approach is based on modified fractal signature. Instead of the time-consuming traditional approaches (top-down and bottom-up approaches) where iterative operations are necessary to break a document into blocks to extract its geometric (layout) structure, this new approach can divide a document into blocks in only one step. This approach can be used to process documents with high geometrical complexity. Experiments have been conducted to prove the proposed new approach for document processing.

Based on wavelets, a new theoretical method has been developed to process form documents. In this method, two-dimensional multiresolution analysis (MSA), wavelet decomposition algorithm, and compactly supported orthonormal wavelets are used to transform a document image into sub-images. According to these sub-images, the reference lines of forms can be extracted, and knowledge about the geometric structure of the document can be acquired. Experiments prove that this new method can be applied to process documents with promising results.

In this paper, a simple and robust algorithm is presented to detect skew in a totally unconstrained document. It can discover the skew angle not only in the whole page of document but also in different document blocks which have their different skew angles. This method consists of four major phases, namely: (a) skew detection and correction for whole page; (b) segmentation of document into blocks; (c) identification of skewed text blocks, and (d) skew detection and correction for the skewed text blocks. To detect the skew in a document, the saw-tooth algorithm and least squares method are used. To segment a document into blocks, the fractal approach is applied. Promising experimental results are also provided to prove the effectiveness of the proposed method.

A new non-hierarchical spatial data structure named four directional adjacency graphs (FDAG) is proposed. In the FDAG vertical and horizontal neighborhood relationship between rectangles is well represented so that structural information can be easily extracted. An application for structural analysis of forms is given, where experiments are conducted with positive results.

Due to different writing styles and various kinds of noise, the recognition of handwritten numerals is an extremely complicated problem. Recently, a new trend has emerged to tackle this problem by the use of multiple classifiers. This method combines individual classification decisions to derive the final decisions. This is called "Combination of Multiple Classifiers" (CME). In this paper, a novel approach to CME is developed and discussed in detail. It contains two steps: data transformation and data classification. In data transformation, the output values of each classifier are first transformed into a form of likeness measurement. The larger a likeness measurement is, the more probable the corresponding class has the input. In data classification, neural networks have been found very suitable to aggregate the transformed output to produce the final classification decisions. Some strategies for further improving the performance of neural networks have also been proposed in this paper. Experiments with several data transformation functions and data classification approaches have been performed on a large number of handwritten samples. The best result among them is achieved by using both the proposed data transformation function and the multi-layer perceptron neural net, which increased the recognition rate of three individual classifications considerably.

For pattern recognition, when a single classifier cannot provide a decision which is 100 percent correct, multiple classifiers should be able to achieve higher accuracy. This is because group decisions are generally better than any individual's. Based on this concept, a method called the "Behavior-Knowledge Space Method" was developed, which can aggregate the decisions obtained from individual classifiers and derive the best final decisions from the statistical point of view. Experiments on 46451 samples of unconstrained handwritten numerals have shown that this method achieves very promising performances and outperforms voting, Bayesian, and Dempster-Shafer approaches.>

Skeletonization algorithms have played an important role in the preprocessing phase of OCR systems. In this paper we report on the performance of 10 parallel thinning algorithms from this perspective by gathering statistics from their performance on large sets of data and examining the effects of the different thinning algorithms on an OCR system.>

A regional decomposition method is proposed to facilitate pattern analysis and recognition. It splits a complicated pattern into several simple parts or sub-patterns, so that the pattern can be identified by examining the distinct parts. A complexity analysis is derived in this paper to prove the effectiveness of the regional decomposition method; mathematical and statistical formulas are also provided to evaluate the recognition rates of different parts. For a sample of 36 alphanumeric characters handprinted in 89 most common styles, the total mean recognition rates of parts have been found to be 30% higher than those obtained from subjective experiments.>

Millions of financial transactions take place every day. Associated with them are documents such as bank cheques, payment slips and bills which have to be processed. A great deal of time, effort and money will be saved if they can be entered into the computer and processed automatically. According to the specific characteristics of financial documents, it can be concluded that it is possible to build a system for recognizing specific types of financial documents, instead of a complex and general one aiming at different kinds of documents. In this paper, a financial document recognition prototype system which can process bank cheques, payment slips and bills, is presented. It consists of four major parts: (a) document image acquisition including scanning and binarization, (b) fixed document processing subsystem based on the detection of staff lines, (c) flexible document processing subsystem operating in a form description language (FDL), and (d) character recognition. Numerous experimental results are presented and discussed.>

Due to different writing styles and various kinds of noise, the recognition of handwritten numerals is an extremely complicated problem. A new trend to tackle this task by the use of multiple classifiers has emerged, which is called "combination of multiple classifiers" (CME). In this paper, a novel approach for CME is developed and discussed in detail. It contains two steps: data transformation and data classification. In data transformation, the output values of each classifier are first transformed into a form of likeness measurement. In data classification, neural-networks have been found very suitable to aggregate the transformed output and produce the final classification decisions. Experiments on 46,451 handwritten numerals have shown a great improvement in recognition by using the present method.

Recently, it has been demonstrated that combining the decisions of several classifiers can lead to improved recognition results. The combination can be implemented using a variety of strategies, among which majority vote is by far the simplest, yet it has been found to be just as effective as more complicated schemes. However, all the results reported thus far on combinations of classifiers have been experimental in nature. The intention of this research is to analyze the foundations of the majority vote method in order to gain a deeper understanding and new results about its mode of operation.

One of the most important topics in handwritten character recognition is the extraction of features from character images. In this paper, an algebraic feature extraction technique is applied to recognize handwritten characters. The discriminant performance of the algebraic features extracted from both handprinted characters and totally unconstrained handwritten numerals is studied. Experimental results are provided.

An efficient sequential method of tracing contour chains around foreground objects in an image is presented. The image is scanned row by row, extracting the run lengths of each row, determining their connectivity with runs in the preceding row, and incrementally building up contour chains accordingly. The contours are stored in a new list structure that reflects their topological nesting in the image. The organization of this list provides a greater degree of topological information than is provided by other methods.

The covariance analysis of linear predictive coding has wide applications, especially in speech recognition and speech signal processing. Real-time applications demand very high processing speed for linear predictive coding analysis. VLSI technology which possesses properties of low-cost, high-speed and massive computing capabilities is a suitable candidate. In this paper, systolic array processors for the covariance analysis of linear predictive coding are developed. The covariance analysis of linear predictive coding contains a large set of irregular and nested recurrence equations. Systolizing the algorithm is a difficult task for such a complex problem, Existing methods of systematic design for systolic arrays are not much helpful to this problem. To overcome it, a break-combination method is presented in this paper. In this manner, the task is first decomposed and then mapped onto several interconnected systolic arrays. The resulting systolic arrays of the sub problems are then combined to form a complete solution.

Extraction of a stable and representative set of features is the heart of the design of a pattern recognition system. Knowing the distribution of information on the pixels of a character will be of great assistance to the study of feature extraction. In this paper, an analysis of the distribution of information on the pixels of binarized Chinese characters is presented. From the analysis, it is obvious that the information of a Chinese character tends to concentrate around the peripheries of the character. Several methods to extract peripheral shape features are presented. Some experiments are conducted on Chinese character recognition and the results show the advantages of the peripheral shape features.

The increasing application of rule-based expert systems has led to the urgent need to quantitatively measure their quality, especially the maintainability which is harder and more expensive than that of conventional software because of the dynamic and evolutionary features of rules. One of the main factors that affect the maintainability of rule-based expert systems is their complexity; but so far little effort has been devoted to measure it. The paper investigates several complexity metrics for rule-based expert systems, and presents some evaluation methods based on statistical testing, analysis and comparison to assess the validity of these metrics. 71 rule-based expert systems are collected as test data from different application areas. The results reveal that a properly defined complexity metric, like our proposed RC, can be used as an effective means to measure the complexity of rule-based expert systems.>

Knowing the structure of a document is the key to successful processing of a document. There exist a variety of definitions of document structures. This paper is a survey of methods describing document structures. Several novel concepts and theoretical analyses are also presented. A document not only has a concrete two-dimensional image but also a conceptual structure which corresponds to human’s thinking. The process of publishing or writing corresponds to encoding the conceptual structure into a concrete structure. Conversely, the concrete structure of the document is decoded into its conceptual one in document processing. In this paper, conceptual and concrete structures are introduced. A complete system for treating both of the conceptual and concrete structures is probably still decades away. As the first stage, this study puts some emphasis on concrete structures, for which, geometric, logical, textual, information, textural, and other structures are described.

This paper describes a new approach that leads to the discovery of substitutions or approximations for physical transformation by fixed and elastic geometric transformation models. These substitutions and approximations can simplify the solution of normalization and generation of shapes in signal processing, image processing, computer vision, computer graphics, and pattern recognition. In this paper, several new algorithms for fixed geometric transformation models such as bilinear, quadratic, bi-quadratic, cubic, and bi-cubic are presented based on the finite element theory. To tackle more general and more complicated problems, elastic geometric transformation models including Coons, harmonic, and general elastic models are discussed. Several useful algorithms are also presented in this paper. The performance of the proposed approach has been evaluated by a series of experiments with interesting results.

The knowledge acquisition bottleneck has become the major impediment to the development and application of effective information systems. To remove this bottleneck, new document processing techniques must be introduced to automatically acquire knowledge from various types of documents. By presenting a survey on the techniques and problems involved, this paper aims at serving as a catalyst to stimulate research in automatic knowledge acquisition through document processing. In this study, a document is considered to have two structures: geometric structure and logical structure. These play a key role in the process of the knowledge acquisition, which can be viewed as a process of acquiring the above structures. Extracting the geometric structure from a document refers to document analysis; mapping the geometric structure into logical structure is regarded as document understanding. Both areas are described in this paper, and the basic concept of document structure and its measurement based on entropy analysis is introduced. Logical structure and geometric models are proposed. Both top-down and bottom-up approaches and their entropy analyses are presented. Different techniques are discussed with practical examples. Mapping methods, such as tree transformation, document formatting knowledge and document format description language, are described.>

In this paper, an advanced hierarchical model has been proposed to produce a more effective character recognizer based on the probability of occurrence of the patterns. New definitions such as crucial parts, efficiency ratios, degree of confusion, similar character pairs, etc. are also given to facilitate pattern analysis and character recognition. Using these definitions, computer algorithms have been developed to recognize the characters by parts, including halves, quarters, and sixths. The recognition rates have been analyzed and compared to those obtained from subjective experiments. Based on the results of both computer and human experiments, a detailed analysis of the crucial parts and the Canadian standard alphanumeric character set has been made which revealed some fundamental characteristics of these handprint models. The results should be useful to pattern analysis and recognition, character understanding, handwriting education, and human-computer communication.>

This paper presents the regional projection contour transformation (RPCT) which transforms a compound pattern or multicontour pattern into a unique outer contour. Two RPCT's, (1) diagonal-diagonal regional projection contour transformation and (2) horizontal-vertical regional projection contour transformation, are presented. They are applicable to a wide range of areas such as image analysis, pattern recognition, etc. A very large scale integration (VLSI) architecture to implement the RPCT has also been designed based on a canonical methodology which maps homogeneous dependence graphs into processor arrays. In this paper, a linear array has been designed, where an N/2-element vector is used to process a pattern with a size of N/spl times/N. It can speed up the recognition process considerably with a time complexity of O(N) compared with O(N/sup 2/) when a uniprocessor is used.>

A cheque processing system currently under development is described. More precisely, the cursive script recognition module for the legal amount is discussed. Commonly, systems perform recognition either on a character by character basis, or on a word level. The authors investigate the recognition at a higher level of abstraction, at the sentence level. Knowledge of context, orthography, syntax and semantics is used to supplement the information from the graphical input. The system is based on a psychological model of the reading process for a fast reader. The module for extracting graphical clues, implemented with the techniques of mathematical morphology, is discussed.>

An approach for the combination of classifiers, in the context that each classifier can offer not only class labels but also the corresponding measurement values, is introduced. This approach is called the Linear Confidence Accumulation method (LCA). The three steps that LCA consists of are: first, measurement values; second, a confidence aggregation function aggregates the confidence values of each class label; and the last, the final decision will be derived by a decision rule based on the accumulated confidence values. Preliminary experiments have been performed and showed that LCA achieved better performance than the voting and the Bayesian methods. This reveals that measurement values play an important role in improving a system's performance when combining different classifiers.>

The authors report on the performance of 10 parallel thinning algorithms from the perspective of character recognition. The algorithms evaluated include the complete range of four-subcycle, two-subcycle, and the most recent fully parallel methods. The authors consider different aspects of the performance of each algorithm. Statistics are gathered such as computing time, deviation from perfect 8-connectedness, and number of possible noise spurs present in the skeletons. In addition, the effects of each algorithm on an OCR system are examined through training and testing the system on the skeletons obtained from each thinning process.>

A method for comparing thinning algorithms involving a series of experiments with human subjects is presented. Several statistical tests are reported to analyze the preference structure exhibited by the data. It is concluded that humans are coherent in comparing thinning algorithm outputs and that reference skeletons can be useful to facilitate the evaluation of thinning algorithms. An evaluation protocol is then proposed and applied to the design and evaluation of a new thinning algorithm for handwritten character recognition.>

A new method of separating touching unconstrained handwritten digits is proposed. A binary image containing a string of touching digits is scanned to give contour chains. The chains are analyzed and subdivided into four kinds of regions: valleys, mountains, holes, and open regions. Individual points of interest in the outer contour are then identified, e.g., points of high curvature. The separating path is assumed to pass between some pair of these significant contour points (SCPs). Nine features of the SCPs are measured and are used to sort the list of all possible pairings of SCPs. Preliminary results show that the correct cut is sorted within the first three choices in 89% of tests.>

Knowing the structure of a document is the key to successful processing of that document. From different points of view, there exist different definitions for document structures. A survey which contains a collection of many methods of describing document structures is presented. Several novel concepts and theoretical analyses are also presented in this survey.>

An intelligent document processing system which can process unconstrained format-free documents is presented. A new document format definition language called Document Architecture Language (DAL) which can handle both rectangular and irregular blocks is presented. Combined with pattern recognition techniques, DAL can be widely used in document analysis and understanding with very wide scopes including general newspaper articles, editorials, financial forms, advertisements, front page of magazine, etc. Bank cheques are employed as an example.>

A new thinning algorithm based on the well known concept of the force of attraction or repulsion between charged particles is presented. This algorithm generates connected skeletons which preserve the shape and end-points of the original patterns. Its performance is experimentally compared with four other known algorithms published in the literature. For the sake of comparison, a reasoned set of test data is introduced. The results of our comparison reveal that the proposed CPM (Charge Particle Method) algorithm is almost as fast as the fastest of those compared. For thin images obtainable from low resolution scanners or coarse scanned images, the CPM algorithm is the fastest.

Whilst the design of skeletonization algorithms has been a very active research area, methodologies for an automatic evaluation of the quality of the results remain to be developed. The difficulty rests on the fact that certain geometric properties considered desirable in skeletons (especially for pattern recognition applications) are not easily quantifiable by means other than human vision. The attempt here is to develop distance measures based on shape matching methods, and to compare skeletons to references in terms of these distances. Three such methods have been adapted for this purpose, and the results are found to be highly correlated for the samples tested.

As a result of its central role in the preprocessing of image patterns, or because of its intrinsic appeal, the design of skeletonization algorithms has been a very active research area. However, few attempts have been made to evaluate the performance of different skeletonization algorithms. This paper presents the results of experiments to evaluate the performance of 20 skeletonization algorithms previously published in the literature. These algorithms have been implemented on the SUN 3/60 workstation in C and tested with a large variety of character patterns. A systematic comparison of these algorithms has been made based on the following criteria: reconstructibility, computation speed, similarity to the reference skeleton, quality of the skeleton, connectivity after skeletonization, and the degree of parallelism.

This paper investigates three different methods of comparing preference structures for thinning algorithms. The first method involves a series of experiments with human subjects. The second makes use of neural networks and the third is based on dissimilarities and distance measures that is computer generated. Several statistical tests have been performed to analyze the preference structures exhibited by the data. This study highlights human coherence in comparing skeletons and the novelty of using reference skeletons to facilitate the evaluation of thinning algorithms. None of the automatic approaches provides a useful insight although a measure of information content manifests some consistency. The overall study suggests a systematic protocol involving human coherence to evaluate preprocessing algorithms.

Nonlinear shape distortions are considered as uncertainty in computer vision, robot vision, and pattern recognition. A new approach to nonlinear shape restoration based on nonlinear image shape transformation is proposed. The principal idea of this method is that two-dimensional (2-D) transformation is used to approximate a three-dimensional (3-D) problem. Five particular image transformation models, bilinear, quadratic, cubic, biquadratic, and bicubic models, are presented in this paper to handle some special cases. Two general transformation models, Coons and harmonic models, are also introduced to tackle more general and more complicated problems. These models are derived from finite-element theory and they can be used to approximate some nonlinear shape distortions under certain conditions. Furthermore, their inverse transformations can be used to remove nonlinear shape distortions. Some useful algorithms are developed. The performance of the proposed approach for nonlinear shape restoration has been evaluated in several experiments with interesting results.>

The authors propose a character segmentation algorithm which is a region-based approach using background pixels. Some interesting background regions are found automatically by performing two independent filtering steps: top-down filtering and bottom-up filtering. Relationships between the resulting background components are considered to select plausible regions favorable to segmentation. For preliminary experimentation, the authors collected 120 samples of connected digits written by 12 people, 97 pairs were successfully segmented.>

While the design of skeletonization algorithms has been a very active research area, methodologies for an automatic evaluation of the quality of the results remain to be developed. The attempt here is to develop distance measures based on shape matching methods, and to compare skeletons references in terms of these distances. Three such methods have been adapted for this purposes, and the results are found to be highly correlated for the samples tested.>

Examines the extraction of curvature features from the contours of 2D objects. General schemes for corner detection and particular methods from the recent OCR literature are considered. Eight approaches are compared with respect to their results on a set of 100 handwritten numerals of varying styles and sizes. Strengths and weaknesses are summarized for each method.>

The number of Chinese characters is very large, frequently it exceeds 5000 in daily usage. In order to achieve accuracy and speed in the recognition of handwritten Chinese characters, it is essential to have a well-organized model database. In the paper, the structural and statistical information of Chinese characters are represented by hierarchical attributed graphs. A heterogeneous multiway tree structure is used to organize the model database. For an input character, a corresponding model character in the database is found by a search process which can be divided into a number of simple and local decisions at different levels of the tree. The matching process becomes quite efficient and accurate.>

An advanced hierarchical model has been proposed to produce a more effective character recognizer based on the probability of occurrence of the patterns. New definitions such as crucial parts, efficiency ratios, degree of confusion, similar character pairs, etc. have also been given to facilitate pattern analysis and character recognition. Using these definitions, computer algorithms have been developed to recognize the characters by parts, including halves, quarters, and sixths. The recognition rates have been analyzed and compared with those obtained from subjective experiments. Based on the results of both computer and human experiments, a detailed analysis of the crucial parts and the Canadian standard alphanumeric character set has been made revealing some interesting fundamental characteristics of these handprint models. The results should be useful for pattern analysis and recognition, character understanding, handwriting education, and human-computer communication.>

A method called concentration-contour method is presented. It transforms a compound pattern into an integral one where contour analysis can be used. The concentration-contour method consists of four phases: (1) concentration of pattern, (2) extraction of contour, (3) transformation of numerical features, and (4) classification. The diagonal-diagonal regional projection transformation (DDRPT), which converts a compound pattern into an integral object, has been used. A VLSI architecture to implement the concentration-contour approach has been designed. The time complexity of the method is only O(N) compared with O(N/sup 2/) when a uniprocessor is used.>

Presents a new approach called regional projection transformation (APT) which converts a compound pattern into an integral object. Diagonal-diagonal regional projection transformation (DDRPT), has been described and analyzed. The patterns transformed from this method possesses a couple of important characteristics which facilitate the recognition of compound patterns. Parallel algorithm for the DDRPT has been presented in this paper. It can speed up computation and the recognition process.>

An optimal algorithm for detecting straight lines in chain codes is described. The algorithm turns the complicated problem of determining the straightness of digital arcs into a simple task by constructing a passing area around the pixels. It is shown that this algorithm is not only simple and intuitive, but also highly efficient.>

An optimal pairing scheme in associative memory classifier is discussed, especially when Hadamard vectors are selected as inner codes. As such a scheme is applied to recognize a set of multi-font Chinese characters, an improvement in the classification performance of this associative memory network is observed.>

A comprehensive survey of thinning methodologies is presented. A wide range of thinning algorithms, including iterative deletion of pixels and nonpixel-based methods, is covered. Skeletonization algorithms based on medial axis and other distance transforms are not considered. An overview of the iterative thinning process and the pixel-deletion criteria needed to preserve the connectivity of the image pattern is given first. Thinning algorithms are then considered in terms of these criteria and their modes of operation. Nonpixel-based methods that usually produce a center line of the pattern directly in one pass without examining all the individual pixels are discussed. The algorithms are considered in great detail and scope, and the relationships among them are explored.>

The splitting-integrating method is a technique developed for the normalization of images by inverse transformation. It does not require solving nonlinear algebraic equations and is much simpler than any existing algorithm for the inverse nonlinear transformation. Moreover, its solutions have a high order of convergence, and the images obtained through T/sup -1/ are free from superfluous holes and blanks, which often occur in transforming digitized images by other approaches. Application of the splitting-integrating method can be extended to supersampling in computer graphics, such as picture transformations by antialiasing, inverse nonlinear mapping, etc.>

Research and development of OCR systems are considered from a historical point of view. The historical development of commercial systems is included. Both template matching and structure analysis approaches to R&D are considered. It is noted that the two approaches are coming closer and tending to merge. Commercial products are divided into three generations, for each of which some representative OCR systems are chosen and described in some detail. Some comments are made on recent techniques applied to OCR, such as expert systems and neural networks, and some open problems are indicated. The authors' views and hopes regarding future trends are presented.>

Four independently, developed expert algorithms for recognizing unconstrained handwritten numerals are presented. All have high recognition rates. Different experimental approaches for incorporating these recognition methods into a more powerful system are also presented. The resulting multiple-expert system proves that the consensus of these methods tends to compensate for individual weaknesses, while preserving individual strengths. It is shown that it is possible to reduce the substitution rate to a desired level while maintaining a fairly high recognition rate in the classification of totally unconstrained handwritten ZIP code numerals. If reliability is of the utmost importance, substitutions can be avoided completely (reliability=100%) while retaining a recognition rate above 90%. Results are compared with those for some of the most effective numeral recognition systems found in the literature.>

Possible solutions to the problem of combining classifiers can be divided into three categories according to the levels of information available from the various classifiers. Four approaches based on different methodologies are proposed for solving this problem. One is suitable for combining individual classifiers such as Bayesian, k-nearest-neighbor, and various distance classifiers. The other three could be used for combining any kind of individual classifiers. On applying these methods to combine several classifiers for recognizing totally unconstrained handwritten numerals, the experimental results show that the performance of individual classifiers can be improved significantly. For example, on the US zipcode database, 98.9% recognition with 0.90% substitution and 0.2% rejection can be obtained, as well as high reliability with 95% recognition, 0% substitution, and 5% rejection.>

The size-orientation-invariance characteristic plays an important role in pattern recognition. It has many applications in computer vision, optical character recognition (OCR), office automation, electronic publication, graphics, etc. In this paper, a new method called transformation-ring-projection (TRP) is proposed to achieve this characteristic. In TRP, shape transformation technique is employed to center the pattern image and normalize its size; the ring-projection scheme is used to handle the orientation problem. An experiment was conducted to verify the proposed method in character recognition. The TRP algorithm requires only simple and regular operations, and provides the feasibility or VLSI implementation to speed up computation for real-time processing. A study on VLSI architecture with extensive parallel processing and pipelining capabilities for the proposed TRP algorithm is also presented.

An image transformation method that performs mapping and filling at the same time, while respecting the connectivity of the original image, is proposed. As a result, the transformations become more consistent and accurate. Its VLSI implementation can reduce the time complexity to O(N/sup 2/) using a uniprocessor, where N is the dimension of the image plane. The algorithms can handle all kinds of images including those of long narrow objects that present problems to other algorithms They also reduce the errors introduced by the order in which rotation and scaling are applied. Their application to gray level images is studied. A series of experiments has been conducted to verify the performance of the proposed algorithms.>

A harmonic model of shape transformations is presented. In this model, harmonic functions are governed by the Laplace equation. This model can convert all image or a pattern to another with arbitrary shapes. The transformation process is harmonic, without abruptness and discontinuity. This model can be used to generate and recognize handwritten Roman letters and Chinese characters, fingerprints, and other types of digitized images and patterns. The algorithms of the harmonic models involve partial differential equations and their numerical solutions. Therefore, an intrinsic link between pattern images and numerical methods is presented.>

Two novel methods for recognizing totally unconstrained handwritten numerals are presented. One classifies samples based on structural features extracted from their skeletons; the other makes use of their contours. Both methods achieve high recognition rates (86.05%, 93.90%) and low substitution rates (2.25%, 1.60%). To take advantage of the inherent complementarity of the two methods, different ways of combining them are studied. It is shown that it is possible to reduce the substitution rate to 0.70%, while the recognition rate remains as high as 92.00% . Furthermore, if reliability is of utmost importance, one can avoid substitutions completely (reliability 100%) and still retain a fairly high recognition rate (84.85%).>

An entropy-reduced transformation (ERT) approach to nonlinear shape restoration has been developed. Nonlinear shape distortions are formulated using nonlinear shape transformations derived from the finite-element theory. Several algorithms which perform the nonlinear shape transformations are given. The inverse nonlinear shape transformation algorithms are described. Some application experiments are described, and results are given.>

A novel recognition system has been implemented to solve the difficult problem of handwritten numeral recognition. In this system, the Fourier descriptors are used as dominant features, and a modified backpropagation model is applied to classification. A novel backpropagation learning algorithm has been developed, and its performance has been evaluated. The results show that the learning algorithm is superior to the original backpropagation model. The proposed algorithm was able to solve the nonconvergence problem typically occurring with the standard backpropagation approach. The algorithm has been tested on handwritten numerals collected by the US Post Office

This paper presents linear and bilinear shape transformations including basic transformations, analyzes their geometric properties, and provides computer algorithms. The shape transformations can be used to simplify the recognition of Roman letters, Chinese characters and other pictorial patterns by normalizing their shapes to the standard forms. Important theoretical analyses have been performed to illustrate that the linear and bilinear transformations are applicable to computer recognition of digitized patterns. A number of pictorial examples have been computed to confirm the analyses and conclusions made.

New splitting-shooting methods are presented for nonlinear transformations T: ( xi , eta ) to (x,y) where x=x( xi , eta ), y=y( xi , eta ). These transformations are important in computer vision, image processing, pattern recognition, and shape transformations in computer graphics. The methods can eliminate superfluous holes or blanks, leading to better images while requiring only modest computer storage and CPU time. The implementation of the proposed algorithms is simple and straightforward. Moreover, these methods can be extended to images with gray levels, to color images, and to three dimensions. They can also be implemented on parallel computers or VLSI circuits. A theoretical analysis proving the convergence of the algorithms and providing error bounds for the resulting images is presented. The complexity of the algorithms is linear. Graphical and numerical experiments are presented to verify the analytical results and to demonstrate the effectiveness of the methods.>

This survey describes the state of the art of online handwriting recognition during a period of renewed activity in the field. It is based on an extensive review of the literature, including journal articles, conference proceedings, and patents. Online versus offline recognition, digitizer technology, and handwriting properties and recognition problems are discussed. Shape recognition algorithms, preprocessing and postprocessing techniques, experimental systems, and commercial products are examined.>

Four nonlinear shape models are presented: polynomial, Coons, perspective, and projective modes. Algorithms and some properties of these models are provided. For a given physical model, such as a perspective model, comparisons are made with other mathematical models. It is proved that, under certain conditions, the perspective models can be replaced by the Coons models. Problems related to substitution and approximation of practical models that facilitate digital image processing are raised and discussed. Experimental results on digital images are presented.>

A method of recognizing unconstrained handwritten numerals using a knowledge base is proposed. Features are collected from a training set and stored in a knowledge base that is used in the recognition stage. Recognition is accomplished by either an inference process or a structural method. The scheme is general, flexible, and applicable to different methods of feature extraction and recognition. By changing the acceptance parameters, a continuous range of performance can be achieved. Encouraging results on nearly 17000 totally unconstrained handwritten numerals are presented. The performance of the system under different recognition-rejection tradeoff ratios is analyzed in detail.>

An algorithm is presented for describing and discriminating binary images of planar shapes. The descriptor is a matrix that the dimensions dependent on the maximum radius of the shape. The descriptor is obtained by a polar quantization of the shape and is independent of the shape's position, orientation, and size. The descriptors discriminate two or more shapes by measuring their degree of similarity. The shapes considered include letters of the alphabet, numerals, geometrical figures, and physical objects. The results obtained for the letters were compared to those reported by A. Goshtasby (1985). The comparison showed that the proposed algorithm provides a higher degree of discrimination.>

Two kinds of Fourier shape descriptors (FDs) are considered: ZR defined by C.T. Zahn and R.S. Roskies (1972) and G defined by G.H. Granlund (1972). In the first part of the paper ZR descriptors are studied. Two modifications of ZR descriptors are proposed. The new descriptors are based on step signature and smoother signature. The amplitudes of FDs are invariant under rotations, translations, changes in size, mirror reflections, and shifts in the starting point. In all the cases the reconstruction accuracy in terms of the number of FDs is studied, resulting in approximation error bounds. An efficient reconstruction method not requiring numerical integration is proposed for polygonal shapes. In the second part of the work theoretical results are verified in numerical experiments involving handwritten characters. In the same experiments, the performances of ZR and G descriptors are compared.>

The state of the art of online handwriting recognition is surveyed based on an extensive review of the literature. Shape recognition algorithms, preprocessing and postprocessing techniques, experimental systems, and commercial products are examined. It is found that online recognizers, except for bar-code readers and other specialized equipment, are not applicable to machine-printed characters.>

An algorithm is presented of extracting features from Chinese characters. These features consist of the Fourier spectrum of projections obtained from multiple-layers of annular partitions. This method takes into consideration the square shape of Chinese characters to that the extracted features contain the significant information of the different parts of the character, and are insensitive to rotation and linear displacement. For the experiments, 97 similar Chinese characters were selected from the most frequently used characters. These characters were divided into 34 groups according to similarity in shape. Three different fonts of Chinese characters (Song, Kai and Bold face) were used. Four additional symbols were also included to study the effects of character symmetry on the proposed algorithm. Experimental results indicate that for any displacement and for rotations in the range of (-180 degrees , +180 degrees ), this method can separate without exception all similar Chinese characters including the complex ones.>

This paper deals with the application of automatic sorting of envelopes with totally unconstrained handwritten numeric postal ZIP codes and presents a complete model (including preprocessing, feature extraction and classification modules) of a ZIP code reader/sorter. Different recognition methods, including statistical, structural and combined were developed and their performance on real-life ZIP code samples (8540 numerals) were measured. The statistical recognition method was used as a front-end recognizer and predictor of an unknown character. Based on edge classification, a new technique was implemented to define and extract the structural features. In the combined recognition method, unknown characters were identified either by the statistical or structural method. Its recognition reliability was found to be in the interval (96.29%, 95.94%) with substitution and rejection rates between (3.45%, 3.96%) and (2.36%, 7.01%) respectively.

In the tree classifier with top-down search, a global decision is made via a series of local decisions. Although this approach gains in classification efficiency, it also gives rise to error accumulation which can be very harmful when the number of classes is very large. To overcome this difficulty, a new tree classifier with the following characteristics is proposed: 1) fuzzy logic search is used to find all ``possible correct classes,'' and some similarity measures are used to determine the ``most probable class''; 2) global training is applied to generate extended terminals in order to enhance the recognition rate; 3) both the training and search algorithms have been given a lot of flexibility, to provide tradeoffs between error and rejection rates, and between the recognition rate and speed. A computer simulation of the decision trees for the recognition of 3200 Chinese character categories yielded a very high recognition rate of 99.93 percent and a very high speed of 861 samples/s, when the program was written in a high level language and run on a large multiuser time-sharing computer.

Based on a recursive process of reducing the entropy, the general decision tree classifier with overlap has been analyzed. Several theorems have been proposed and proved. When the number of pattern classes is very large, the theorems can reveal both the advantages of a tree classifier and the main difficulties in its implementation. Suppose H is Shannon's entropy measure of the given problem. The theoretical results indicate that the tree searching time can be minimized to the order O(H), but the error rate is also in the same order O(H) due to error accumulation. However, the memory requirement is in the order 0(H exp(H)) which poses serious problems in the implementation of a tree classifier for a large number of classes. To solve these problems, several theorems related to the bounds on the search time, error rate, memory requirement and overlap factor in the design of a decision tree have been proposed and some principles have been established to analyze the behaviors of the decision tree. When applied to classify sets of 64, 450, and 3200 Chinese characters, respectively, the experimental results support the theoretical predictions. For 3200 classes, a very high recognition rate of 99.88 percent was achieved at a high speed of 873 samples/s when the experiment was conducted on a Cyber 172 computer using a high-level language.

A multistage classifier with general tree structure has been developed to recognize a large number of Chinese characters. A simple and efficient method of classifying the characters was achieved by choosing the best feature at each stage of the tree. The features used are Walsh coefficients obtained from two profiles of a character projected onto the X-Y orthogonal axes. Some algorithms for aligning the characters were compared and one of them was adopted in this recognition scheme. A high recognition rate of about 99.5 percent was obtained in an experiment with more than 3000 different Chinese characters.

Since handwritten characters vary in shape and writing-stroke sequence, it is desirable to develop a standard set of characters that are of high quality, so that not only are they easy to write, but they are also most suitable for machine recognition. A database of more than 100 000 alphanumeric patterns was assembled. It consisted of 174 models of the alphanumeric characters written by both left-handed and right-handed subjects. Based on frequency density and distance measurements, a nietric called the dispersion factor was computed to rank the various models. The principle of the metric is discussed, and results are given indicating the high quality models of the alphanumerics.

n-gram (n = 1 to 5) statistics and other properties of the English language were derived for applications in natural language understanding and text processing. They were computed from a well-known corpus composed of 1 million word samples. Similar properties were also derived from the most frequent 1000 words of three other corpuses. The positional distributions of n-grams obtained in the present study are discussed. Statistical studies on word length and trends of n-gram frequencies versus vocabulary are presented. In addition to a survey of n-gram statistics found in the literature, a collection of n-gram statistics obtained by other researchers is reviewed and compared.