Optical Coherence Tomography (OCT) is a widely used non-invasive imaging technique that provides detailed three-dimensional views of the retina, which are essential for the early and accurate diagnosis of ocular diseases. Consequently, OCT image analysis and processing have emerged as key research areas in biomedical imaging. However, acquiring paired datasets of clean and real-world noisy OCT images for supervised denoising models remains a formidable challenge due to intrinsic speckle noise and practical constraints in clinical imaging environments.To address these issues, we propose SDPA++: A General Framework for Self-Supervised Denoising with Patch Aggregation. Our novel approach leverages only noisy OCT images by first generating pseudo-ground-truth images through self-fusion and self-supervised denoising. These refined images then serve as targets to train an ensemble of denoising models using a patch-based strategy that effectively enhances image clarity. Performance improvements are validated via metrics such as Contrast-to-Noise Ratio (CNR), Mean Square Ratio (MSR), Texture Preservation (TP), and Edge Preservation (EP) on the real-world dataset from the IEEE SPS Video and Image Processing Cup. Notably, the VIP Cup dataset contains only real-world noisy OCT images without clean references, highlighting our method’s potential for improving image quality and diagnostic outcomes in clinical practice.

Handwritten Japanese text recognition is still a big challenging task due to the large character set, diversity of writing styles, and multiple-touches between characters. In this paper, we propose a model of Attention Augmented Convolutional Recurrent Network (AACRN) for recognizing handwritten Japanese text lines. The AACRN model has three main parts: a convolutional feature extractor, a self-attention based encoder, and a CTC-decoder. The whole model can be trained end-to-end. In the experiment, we evaluate the performance of the AACRN model on the TUAT Kondate dataset and the Kuzushiji dataset. The results of the experiments show that the proposed model achieves higher performance than the state-of-the-art recognition accuracies on the test set of TUAT Kondate and the Kuzushiji dataset.

This paper presents an approach for matching various sizes of excavated wooden fragments based on siamese neural networks. We propose a siamese network composed of global average pooling and a fine-tuned Resnet encoder (GA-S-net). We also propose an elaborated siamese network by replacing the global average pooling with a spatial pyramid pooling layer and add a new dense absolute difference layer (SP-S-net). Samples of 37,760 fragments were prepared from 268 complete wooden tablets excavated from the Heijo-Kyo Palace ruins used during the Nara period in Japan. Both of the networks answer whether two fragments are from the same tablet or not. The result of both networks for the testing set is similar to AUC (Area under the curve) of ROC (Receiver Operating Characteristic) curve being around 90%. In AUC of large fragments, however, SP-S-net is better than GA-S-net with 97.1% versus 93.8%. These networks are rather new for dealing with various sizes of inputs for the matching problem.

Recently, the segmentation-free approach using Convolution Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) for handwritten text recognition has been investigated by many research groups. Although good results are produced on some datasets, there remain some drawbacks. It is neither robust against the change of handwriting styles such as gaps between characters, stroke widths and shapes of characters nor stable for skewed, and curved text lines unless heavily trained by such patterns. In this paper, we propose a segmentation-based method for handwritten Japanese text recognition. The method employs a semantic segmentation model for precisely splitting text lines into single characters. The semantic segmentation model is based on an encoder-decoder architecture like U-Net, but we employ available techniques to improve the accuracy of pixel classification. They are a deeper encoder with ResNet 101, dilated convolutions and Spatial Pyramid Pooling. Subsequently, a CNN based OCR is used to recognize segmented character images. Finally, the recognized candidates are considered in a lattice diagram combined with a linguistic context. The result of experiments shows that the segmentation model increases the mean IoU significantly from 89.32% to 94.96% and the proposed approach is robust to the change of handwriting styles while a segmentation-free method is quite sensitive to them, resulting in the significant reduction of the recognition rate.

Discriminating ambiguous symbols in online handwritten mathematical expression is difficult without context. We propose a Bidirectional Recurrent Neural Network for segmenting and classifying online handwritten mathematical symbols. The context from forward and backward directions helps the classification model discriminate ambiguous symbols and improve recognition rates. The classification model is integrated into the Stochastic Context-Free Grammar recognition system for recognizing mathematical expressions. We show the effectiveness of the approach for improving symbol classification and segmentation on the CROHME 2016 dataset.

This paper presents an improvement in recognizing offline handwritten mathematical expressions (HMEs) by deep neural networks. We train it end-to-end using weakly supervised learning. The network has three parts: an encoder using a Convolutional Neural Network to encode high-level features from an input HME image; a decoder using gated recurrent units with attention to parse the high-level features and generate an output expression in the LaTeX format; and a symbol classifier to improve the localization and classification of the high-level features. Besides, we use the model ensemble method for the beam search process to average the probabilities from multiple models. For the dataset of the Competition on Recognition of Online Handwritten Mathematical Expressions (CROHME) 2014 and 2016, we have achieved an expression recognition rate of 53.65% and 51.96% correspondingly, which is 6 points better than without weakly supervised learning. Furthermore, when ensembling several models, the recognition rate of our method is increased to 55.68% for the CROHME 2014 testing set.

We propose a deep neural network-based method to recover dynamic online trajectories from offline handwritten Japanese kanji character images. It is a challenging task since Japanese kanji characters consist of multiple strokes. Our proposed model has three main components: Convolutional Neural Network-based encoder, Long Short-Term Memory Network-based decoder with an attention layer, and Gaussian Mixture Model (GMM). The encoder focuses on feature extraction while the decoder refers to the extracted features and generates time-sequences of GMM parameters. The attention layer is the key component for trajectory recovery. The GMM provides robustness to style variations so that the proposed model does not overfit to training samples. In the experiments, the proposed method is evaluated by both visual verification and handwritten character recognition. This is the first attempt to use online recovered trajectories to help improve offline handwriting recognition performance. Although the visual verification reveals some problems, the recognition experiments demonstrate the effect of trajectory recovery in improving offline handwritten character recognition accuracy when online recognition of the recovered trajectories are combined.

This paper presents an attention-based convolutional sequence to sequence (ACseq2seq) model for recognizing an input image of multiple text lines from Japanese historical documents without explicit segmentation of lines. The recognition system has three main parts: a feature extractor using Convolutional Neural Network (CNN) to extract a feature sequence from an input image; an encoder employing bidirectional Long Short-Term Memory (BLSTM) to encode the feature sequence; and a decoder using a unidirectional LSTM with the attention mechanism to generate the final target text based on the attended pertinent features. We also introduce a residual LSTM network between the attention vector and softmax layer in the decoder. The system can be trained end-to-end by a standard cross-entropy loss function. In the experiment, we evaluate the performance of the ACseq2seq model on the anomalously deformed Kana datasets in the PRMU contest. The results of the experiments show that our proposed model achieves higher recognition accuracy than the state-of-the-art recognition methods on the anomalously deformed Kana datasets.

Despite of recent breakthroughs in the accuracy of single character recognition using the deeper convolution neural networks, one of the remaining problems is that OCRs almost fail to recognize character patterns when they are severely degraded, especially those of the historical documents. Another problem to recognize characters in historical documents is the lack of sufficient training patterns because of the heavy cost for annotation. This paper proposes a character attention generative adversarial network named CAGAN for restoring heavily degraded character patterns in historical documents so that OCRs improve their accuracy and even help archeologists to decode them. The network is based on the U-Net like architecture [1] with skip connections, and it is trained by the proposed loss function including the common adversarial loss (global loss) and the hierarchical character attentive loss (local loss). We made an experiment on 118 categories of most common Japanese Kanji characters, collected from severely damaged historical documents called Heijokyo mokkan written during the Nara period in Japan. The experiment shows that our method restores the shapes of characters and improves the recognition rate significantly, which is helpful for archeologists to decode damaged character patterns.

This paper presents an end-to-end model of Deep Convolutional Recurrent Network (DCRN) for recognizing offline handwritten Japanese text lines. The end-to-end DCRN model has three parts: a convolutional feature extractor using Deep Convolutional Neural Network (DCNN) to extract a feature sequence from a text line image; recurrent layers employing a Deep Bidirectional LSTM to predict pre-frame from the feature sequence; and a transcription layer using Connectionist Temporal Classification (CTC) to convert the pre-frame predictions into the label sequence. Since our end-to-end model requires a large data for training, we synthesize handwritten text line images from sentences in corpora and handwritten character patterns in the Nakayosi and Kuchibue database with elastic distortions. In the experiment, we evaluate the performance of the end-to-end model and the effectiveness of the synthetic data generation method on the test set of the TUAT Kondate database. The results of the experiments show that our end-to-end model achieves higher than the state-of-the-art recognition accuracy on the test set of TUAT Kondate with 96.35% and 98.05% character level recognition accuracies without and with the generated synthetic data, respectively.

This paper presents an attempt to recognize online isolated handwritten Japanese characters as well as text lines by Recurrent Neural Networks (RNNs). Although there are some successful studies on online Chinese handwriting recognition using RNNs, it is difficult to achieve high accuracy for Japanese due to many different types of characters such as kanji and kana. Moreover, training RNNs usually requires a large number of samples for each class of character. Hence, we apply five different transformation operations on the original samples to generate artificial samples with various deformations. For online handwritten Japanese text, we use the Kondate database, but it does not cover the whole Japanese character set. Thus, we generate random text lines using the sentences from corpora and isolated characters from the Nakayosi and Kuchibue databases. Besides, we employ a robust process with some preprocessing steps and the state-of-the-art online features to extract invariant features from handwritten patterns because the handwritten samples are merged from the different databases collected on different devices with various resolutions. For the recognition model, we have implemented the different Bidirectional Long Short-Term Memory Networks (BLSTM networks) for isolated character classification (sequence classification task) and handwritten text recognition (transcription task). The best model of the sequence classification task achieves an accuracy of 97.91% on Nakayosi and 97.74% on Kuchibue. The best model of transcription task performs a character recognition rate of 86.31% on Kondate and 83.15% on the generated text lines.

This paper presents the results of the VOHTR 2018 competition on Vietnamese Online Handwritten Text Recognition. The goal of this competition is to evaluate and compare recent online handwritten text recognition systems on Vietnamese online handwritten text which contains many delayed strokes caused by the diacritic marks. Besides, the general objective is to encourage the studies on Vietnamese online handwritten text recognition based on the large Vietnamese handwriting database collected from 200 writers. In this competition, we introduce three tasks consisting of word recognition (task 1), text-line recognition (task 2) and paragraph recognition (task 3) which are described in details. Subsequently, we describe the evaluation metrics and give comparative results of competitors along with the brief descriptions of the respective methods.

This paper presents our attempts to collect and analyze unconstrained Vietnamese online handwriting text patterns by pen-based computers. Totally, our database contains over 120,000 strokes from more than 140,000 characters, which is one of the largest Vietnamese online handwriting pattern databases currently. For building and analyzing our database, we made a collection tool, a line segmentation tool, and a delayed stroke detection tool. Moreover, we investigated some statistical information from personal information of writers. In order to solve the unconstrained handwriting recognition problem, we conducted experiments using Bidirectional Long Short-Term Memory (BLSTM) networks. BLSTM network is architecture of Recurrent Neural Network (RNN) and applied recently for many related problems. The performance of BLSTM network on our database is nearly 80% of accuracy even though this database contains many delayed strokes. In near future, we are going to avail our database for research purposes, as it would be the fundamental for the handwriting recognition research.

This paper presents a Finite State Machine (FSM) to reduce user's waiting time to get the recognition result after finishing writing in recognition of online handwritten English text. The lexicon is modeled by a FSM, and then determination and minimization are applied to reduce the number of states. The reduction of states in the FSM shortens the waiting time without degrading the recognition accuracy. Moreover, by merging incoming paths to each state, the recognition rate is improved. The N-best states decoding method also reduces the waiting time significantly with small degradation in recognition accuracy. Experiments on IAM-OnDB and IBM_UB_1 show the effectiveness of the method in both reducing waiting and improving recognition accuracy.

This paper presents a semi-incremental recognition method for online handwritten English text. We employ local processing strategy and focus on a recent sequence of strokes defined as "scope". For the latest scope, we build and update a segmentation and recognition candidate lattice and advance the best-path search incrementally. We utilize the result of the best-path search in the previous scope to exclude unnecessary segmentation candidates. This reduces the number of candidate word recognition with the result of reduced processing time. We also reuse the segmentation and recognition candidate lattice in the previous scope for the latest scope. Moreover, triggering recognition processes every few strokes save CPU time. Experiment made on IAM-OnDB database shows the effectiveness of the proposed method not only in reduced processing time and waiting time, but also in recognition accuracy.

This paper presents a semi-incremental recognition method for online Japanese handwritten text recognition, which is used for busy recognition interface (recognition while writing) and lazy recognition interface (recognition after writing) without large waiting time. We employ local processing strategy and focus on a recent sequence of strokes defined as "scope". For the latest scope, we build and update a segmentation and recognition candidate lattice and advance the best-path search incrementally. We utilize the result of the best-path search in the previous scope to exclude unnecessary segmentation candidates. This reduces the number of candidate character recognition with the result of reduced processing time. We also reuse the segmentation and recognition candidate lattice in the previous scope for the latest scope. Moreover, triggering recognition processes every few strokes save CPU time. Experiment made on TUAT-Kondate database shows the effectiveness of the proposed method not only in reduced processing time and waiting time, but also in recognition accuracy.