Reconstruction of skulls from defective models is a very important and challenging task in craniofacial surgery, forensics, and anthropology. Existing methods typically reconstruct approximating surfaces that regard corresponding points on the target skull as soft constraints, thus incurring non-zero error even for non-defective parts and high overall reconstruction error. This paper proposes a novel method that non-rigidly registers an interpolating surface that regards corresponding target points as hard constraints, thus achieving low reconstruction error. To overcome the shortcoming of interpolating surface, a flip-avoiding method is used to detect and exclude conflicting hard constraints that would otherwise cause surface patches to flip. Comprehensive test results show that our method is more accurate than existing methods and it is robust against severe outliers such as radiation artifacts in CT due to dental implants.

Recent advances in robust principle component analysis offers a powerful method for solving a wide variety of low-level vision problems. However, if the input data is very large, especially when high-resolution images are involved, it makes RPCA computationally prohibitive for many real applications. To tackle this problem, we propose a fixed-rank RPCA method that uses coupled dictionaries (FRPCA-CD) to handle high-resolution images. FRPCA-CD downsamples high-resolution images into low-resolution images, performs FRPCA on the low-level images to obtain the low-rank matrix, which is reconstructed at original resolution by coupled dictionaries. Comprehensive tests performed on video background recovery, noise reduction in photometric stereo, and image reflection removal problems show that FRPCA-CD can reduce computation time and memory space drastically without sacrificing accuracy.

Face images are very important in human social activities, which can be severely hampered when they are corrupted by occluders such as eyeglasses, face marks, and scarfs. Existing methods for removing occlusions in face images can be grouped into three broad categories, namely PCA, robust PCA (RPCA), and sparse coding. The major weaknesses of these methods are inconsistent performance across test conditions and possible corruption of unoccluded part of the recovered target image. This paper presents variable-threshold RPCA (VRPCA) method based on RPCA with variable thresholding. Comprehensive tests show that VRPCA is able to preserve the unoccluded parts of the target image with practically zero error. Compared to existing methods, it is more accurate, reliable, and consistent across various test conditions.

Automatic identification of the Frankfurt plane (FP) and mid-sagittal plane (MSP) of a human skull is a very important task. These planes are used to define important landmarks on the skull for surgery planning and guidance, forensic reconstruction, and anthropological and archaeological studies. At present, there is no known method for automatic identification of FP. Existing methods identify FP by fitting a plane over manually marked landmark points. On the other hand, existing automatic methods identify MSP by looking for a symmetric plane that divides the skull into the left and right sides. Unfortunately, this approach is inaccurate because extensive anthropological studies show that there is significant lateral asymmetry in the skulls. This paper proposes an automatic, accurate, and robust method for identifying FP and MSP of skulls. It registers a template skull model with known landmarks to a target skull to locate the landmarks, FP, and MSP on the target skull. Then, it iteratively refines the landmark locations, FP and MSP according to their medical definitions. Test results show that the proposed algorithm is more robust and accurate than symmetry-based methods. Moreover, it can also be applied to partial skull models of real patients.

Studies of interactions between protein domains and ligands are important in many aspects such as cellular signaling and regulation. In this work, we applied a three-stage knowledge-guided approach of docking flexible peptide ligands to SH2 domains. The first stage of the approach search for binding pockets of SH2 domain proteins and binding motifs of peptide ligands based on known features. The knowledge of the binding sites are used in the second stage as binding constraints to align ligand's peptide backbone to the SH2 domain. The backbone-aligned ligands produced serve as good starting points to the third stage which uses a Monte Carlo method to perform the flexible docking. The experimental results show that the backbone alignment method at the second stage produces good backbone conformations which are close to the conformation in complex. The binding site information is well utilized to provide a better starting point to the next docking stage. The subsequent docking is successful or partially successful in 5 of 7 test cases. The performance is better than that of general docking methods. The presented approach can also be applied to other well characterized protein domains which bind ligands in two or more binding grooves.

Abstract The Cosserat theory of elastic rods has been used in a wide range of application domains to model and simulate the elastic deformation of thin rods. It is physically accurate and its implementations are efficient for interactive simulation. However, one requirement of using Cosserat rod theory is that the tubular object must have rigid cross‐sections that are small compared to its length. This requirement make it difficult for the approach to model elastic deformation of rods with large, non‐rigid cross‐sections that can change shape during rod deformation, in particular, hollow tubes. Our approach achieves this task using a hybrid model that binds a mesh elastically to a reference Cosserat rod. The mesh represents the surface of the hollow tube while the reference rod models bending, twisting, shearing and stretching of the tube. The cross‐sections of the tube may take on any arbitrary shape. The binding is established by a mapping between mesh vertices and the rod's directors. Deformation of the elastic tube is accomplished in two phases. First, the reference rod is deformed according to Cosserat theory. Next, the mesh is deformed using Laplacian deformation according to its mapping to the rod and its surface elastic energy. This hybrid approach allows the tube to deform in a physically correct manner in relation to the bending, twisting, shearing, and stretching of the reference rod. It also allows the surface to deform realistically and efficiently according to surface elastic energy and the shape of the reference rod. In this way, the deformation of elastic hollow tubes with large, non‐rigid cross‐sections can be simulated accurately and efficiently.

3D visualization and segmentation of organs in abdominal volume images are important in medical image processing for applications such as diagnosis, treatment and surgical planning. However, the abdominal wall leads to difficulties in both visualization and segmentation. These difficulties can be eliminated by removing the abdominal wall. This paper presents an algorithm that removes abdominal wall by registering a 3D flipping-free deformable model to the inner boundary of the wall. To our best knowledge, it is the first work in removing the abdominal wall for the purpose of 3D visualization and segmentation of the organs.

In this paper, A configuration of Cellular Neural Network (CNN) is introduced to implement Principal Component Analysis (PCA). CNN is a parallel computing paradigm. Many researchers considered it as the next generation universal machine and developed so-called CNN universal chips. Based on the capability of CNN, an alternative PCA implementation named Principal Component Analyzing Cellular Neural Network (PCACNN) is proposed. PCA is used to reduce the dimensions of a given dataset in order to extract the principal information of the given dataset. In decades, many researchers presented their investigations based on PCA in order to improve the performance and/or to attack some open issues in specific fields. In this paper, PCA is implemented based on the architecture and capabilities of CNN. Consequently, the computing performance of PCA can be improved as long as the CNN architecture can be realized.

Medical volume images are large in size. They cannot be efficiently transmitted and visualized as candidates for medical image retrieval and relevance feedback. On the other hand, 2D images that are small in size and rich in 3D details can be efficiently transmitted and visualized as candidates. This paper presents an algorithm that summarizes the 3D details in a volume image into a single 2D image. It applies soft segmentation to highlight the anatomy of interest in the volume, and automatically selects a salient view that contains the most amount of semantic information as the summarization of the volume image. Experimental results show that the proposed method can well summarize medical volume images of different anatomical structures. Compared to representation of volume images using 2D slices and conventional volume rendering, our summarized images are rich in 3D details, and they can be transmitted and visualized very efficiently.

Segmentation of 3D soft organs from complex volume images is a very important and challenging task. The objects of interest may have inhomogeneous voxel intensities and some object boundaries may be indistinct. Existing algorithms are either sensitive to noise or computationally expensive. This paper presents a novel algorithm that overcomes these shortcomings. The algorithm adopts a novel flipping-free mesh deformation and registration method that can easily incorporate geometric constraints to reduce sensitivity to noise. It efficiently deforms the 3D model in large displacements reducing total computational costs. These properties are confirmed by comprehensive test results.

Computer systems are increasingly being used for sports training. Existing sports training systems either require expensive 3D motion capture systems or do not provide intelligent analysis of user’s sports motion. This paper presents a framework for affordable and intelligent sports training systems for general users that require only single camera to record the user’s motion. Sports motion analysis is formulated as a 3D-2D spatiotemporal motion registration problem. A novel algorithm is developed to perform spatiotemporal registration of the expert’s 3D reference motion and a performer’s 2D input video, thereby computing the deviation of the performer’s motion from the expert’s motion. The algorithm can effectively handle ambiguous situations in a single video such as depth ambiguity of body parts and partial occlusion. Test results show that, despite using only single video, the algorithm can compute 3D posture errors that reflect the performer’s actual motion error.

Many large cities have installed surveillance cameras to monitor human activities for security purposes. An important surveillance application is to track the motion of an object of interest, e.g., a car or a human, using one or more cameras, and plot the motion path in a city map. To achieve this goal, it is necessary to localize the cameras in the city map and to determine the correspondence mappings between the positions in the city map and the camera views. Since the view of the city map is roughly orthogonal to the camera views, there are very few common features between the two views for a computer vision algorithm to correctly identify corresponding points automatically. This paper proposes a method for camera localization and position mapping that requires minimum user inputs. Given approximate corresponding points between the city map and a camera view identified by a user, the method computes the orientation and position of the camera in the city map, and determines the mapping between the positions in the city map and the camera view. Both quantitative tests and practical application test have been performed. It can obtain the best-fit solutions even though the user-specified correspondence is inaccurate. The performance of the method is assessed in both quantitative tests and practical application. Quantitative test results show that the method is accurate and robust in camera localization and position mapping. Application test results are very encouraging, showing the usefulness of the method in real applications.

iDVT (interactive Digital Violin Tutor) is a violin learning system exploiting physical and virtual resources and interactivity. It aims at providing the user with new effective learning experience. This demonstration paper briefly describes the structure of the system and the underlying audio-visual processing techniques employed in the system.

Automatic music transcription, in spite of decades of research, remains a challenging research problem. The traditional audio-only approach has yet to achieve a satisfactory performance for any computer-aided pedagogical system. Inspired by the high correlation between violin playing techniques (fingering, bowing) and the played acoustic notes, this paper presents a first attempt in visual analysis of violin fingering to compensate for the difficulties in audio-only music transcription. This is achieved by a robust multiple finger tracking algorithm and a string detection method that extract press, release, and fingertip position from the fingering video and automatically translate the fingering information into the played acoustic note, i.e., onset, offset, and pitches. Experimental results reveal high correctness in multiple finger tracking and string detection, thus paving the way for an improved audio-visual violin transcription system.

Estimation of the 3D pose of bones in 2D images plays an important role in computer-assisted diagnosis and surgery. Existing work has focused on registering a 3D model of the bone to the images of the same bone. This approach incurs a high health care cost on the patients, and are prescribed only for certain procedures. Such computer assistance is not routinely available to diagnostic and surgical procedures that involve only X-ray images, limiting the precision of the procedures. This paper proposes a method for recovering 3D pose by registering a generic 3D model of a bone to single X-ray images of different patients. It can perform 3D-2D registration with unknown correspondence automatically. Comprehensive test results show that the method is accurate and robust in recovering the pose of femurs in the X-ray images of different patients

Removal of rain streaks in video is a challenging problem due to the random spatial distribution and fast motion of rain. This paper presents a new rain removal algorithm that incorporates both temporal and chromatic properties of rain in video. The temporal property states that an image pixel is never always covered by rain throughout the entire video. The chromatic property states that the changes of R, G, and B values of rainaffected pixels are approximately the same. By using both properties, the algorithm can detect and remove rain streaks in both stationary and dynamic scenes taken by stationary cameras. To handle videos taken by moving cameras, the video can be stabilized for rain removal, and destabilized to restore camera motion after rain removal. It can handle both light rain and heavy rain conditions. Experimental results show that the algorithm performs better than existing algorithms.

This paper describes a distributed layered architecture for resource-constrained multirobot cooperation, which is utilized in autonomic mobile sensor network coverage. In the upper layer, a dynamic task allocation scheme self-organizes the robot coalitions to track efficiently across regions. It uses concepts of ant behavior to self-regulate the regional distributions of robots in proportion to that of the moving targets to be tracked in a nonstationary environment. As a result, the adverse effects of task interference between robots are minimized and network coverage is improved. In the lower task execution layer, the robots use self-organizing neural networks to coordinate their target tracking within a region. Both layers employ self-organization techniques, which exhibit autonomic properties such as self-configuring, self-optimizing, self-healing, and self-protecting. Quantitative comparisons with other tracking strategies such as static sensor placements, potential fields, and auction-based negotiation show that our layered approach can provide better coverage, greater robustness to sensor failures, and greater flexibility to respond to environmental changes

3D Human posture sequence estimation from single or multiple image sequences is essential in many applications. However, 3D posture sequence cannot be accurately estimated from single image sequence due to depth ambiguity or self-occlusion, and camera calibration is often required before estimating 3D posture sequence from multiple image sequences. In this paper, we present an algorithm to accurately estimate 3D human posture sequence from two un-calibrated image sequences. The algorithm combines a modified Nonparametric Belief Propagation (mNBP) method with an improved camera self-calibration method. The previously developed mNBP can estimate posture even under partial self-occlusion, and here it is improved to estimate posture when the human model scale is different from that of body image in image sequences. The improved self-calibration can guarantee to find the optimal rotation and relative scale between two fixed but un-calibrated scaled orthographic cameras, without a nonlinear optimization process. Quantitative and qualitative results of experiments show that the algorithm is able to estimate 3D posture sequence from a pair of un-calibrated image sequences. 1

This paper presents our recent work that aims at associating the recognition results of textual and graphical information contained in the scientific chart images. Text components are first located in the input image and then recognized using OCR. On the other hand, the graphical objects are segmented and form high level symbols. Both logical and semantic correspondence between text and graphical symbols are identified. The association of text and graphics allows us to capture the semantic meaning carried by scientific chart images in a more complete way. The result of scientific chart image understanding is presented using XML documents.

In medical applications, sensitivity in detecting medical problems and accuracy of detection are often in conflict. A single classifier usually cannot achieve both high sensitivity and accuracy at the same time. Methods of combining classifiers have been proposed in the literature. This paper presents a study of probabilistic combination methods applied to the detection of bone fractures in X-ray images. Test results show that the effectiveness of a method in improving both accuracy and sensitivity depends on the nature of the method as well as the proportion of positive samples.

Accurate human body posture refinement from single or multiple images is essential in many applications. Two main causes of difficulty to solve the refinement problem are high degree freedom of human body and self-occlusion. One of the most recent algorithms is nonparametric belief propagation (NBP) that solves the problem in a lower dimensional state space. However, it is difficult to handle self-occlusion. This paper presents an NBP-based algorithm that can refine body posture even in self-occlusion case, which has been shown by experimental results. The experimental results also show that our algorithm can accurately refine body posture even if the initial posture has large difference from the true posture.

Retrieval of 3D models have attracted much research interest, and many types of shape features have been proposed. In this paper, we describe a novel approach of combining the feature types for 3D model retrieval and relevance feedback processing.Our approach performs query processing using pre-computed pairwise distances between objects measured according to various feature types. Experimental tests show that this approach performs better than retrieval by individual feature type.

Self-organizing feature maps such as extended Kohonen maps (EKMs) have been very successful at learning sensorimotor control for mobile robot tasks. This letter presents a new ensemble approach, cooperative EKMs with indirect mapping, to achieve complex robot motion. An indirect-mapping EKM self-organizes to map from the sensory input space to the motor control space indirectly via a control parameter space. Quantitative evaluation reveals that indirect mapping can provide finer, smoother, and more efficient motion control than does direct mapping by operating in a continuous, rather than discrete, motor control space. It is also shown to outperform basis function neural networks. Furthermore, training its control parameters with recursive least squares enables faster convergence and better performance compared to gradient descent. The cooperation and competition of multiple self-organized EKMs allow a nonholonomic mobile robot to negotiate unforeseen, concave, closely spaced, and dynamic obstacles. Qualitative and quantitative comparisons with neural network ensembles employing weighted sum reveal that our method can achieve more sophisticated motion tasks even though the weighted-sum ensemble approach also operates in continuous motor control space.

In this paper, we present a novel approach to vectorize elliptic arcs for 3D pie chart recognition and data extraction. As a preprocessing step, a set of straight line segments are obtained first. By estimating the parameters based on the line segments, we are able to generate elliptic arc candidates. Curve fitting is then performed to reject false positives and further refine the parameters of each elliptic arc. The vectorized elliptic arcs are the key components in the 3D pie charts that allow us to determine the chart type and to extract data values embedded. Experiments using a group of testing images show that the proposed method works well and is computationally efficient.

The paper proposes a fuzzy image labeling method that assigns multiple semantic labels together with confidence measures to each region in an image. The confidence measures are derived from the distance of the region to hyperplanes constructed by support vector machines. Test results show that this method yields higher classification accuracy and retrieval precision than crisp labeling methods that are based on crisp classification.

This paper describes a reactive, distributed layered architecture for cooperation of multiple resource-bounded robots, which is utilized in mobile sensor network coverage. In the upper layer, a dynamic task allocation scheme self-organizes the robot coalitions to track efficiently in separate regions. It uses the concepts of ant behavior to self-regulate the regional distributions of robots in proportion to that of the targets to be tracked in the changing environment. As a result, the adverse effects of task interference between robots are minimized and sensor network coverage is improved. In the lower layer, the robots use self-organizing neural networks to coordinate their target tracking within a region. Quantitative comparisons with other tracking strategies such as static sensor placements, potential fields, and auction-based negotiation show that our approach can provide better coverage and greater flexibility in responding to environmental changes.

Most research on automated specification-based software testing has focused on the automated generation of test cases. Before a software system can be tested, it must be set up according to the input requirements of the test cases. This setup process is usually performed manually, especially when testing complex data structures and databases. After the system is properly set up, a test execution tool runs the system according to the test cases and pre-recorded test scripts to obtain the outputs, which are evaluated by a test evaluation tool. This paper complements the current research on automated specification-based testing by proposing a scheme that combines the setup process, test execution, and test validation into a single test program for testing the behavior of object-oriented classes. The test program can be generated automatically given the desired test cases and closed specifications of the classes. With closed specifications, every class method is defined in terms of other methods which are, in turn, defined in their own class specifications. The core of the test program generator is a partial-order planner which plans the sequence of instructions required in the test program. The planner is, in turn, implemented as a tree-search algorithm. It makes function calls to the Omega Calculator library, which solves the constraints given in the test cases. A first-cut implementation of the planner has been completed, which is able to handle simple arithmetics and existential quantifications in the class specifications. A soundness and completeness proof sketch of the planner is also provided in this paper.

Recent advancement in 3D digitization techniques have prompted the need for 3D object retrieval. Our method of comparing 3D objects for retrieval is based on 3D morphing. It computes, for each 3D object, two spatial feature maps that describe the geometry and topology of the surface patches on the object, while preserving the spatial information of the patches in the maps. The feature maps capture the amount of effort required to morph a 3D object into a canonical sphere, without performing explicit 3D morphing. Fourier transforms of the feature maps are used for object comparison so as to achieve invariant retrieval under arbitrary rotation, reflection, and non-uniform scaling o the objects. Experimental results show that our method of retrieving 3D models is very accurate, achieving a precision of above .086 even at a recall rate of 1.0.

Research on digital analysis of painting images has received very little attention. The exact nature of scientific methods seems to be antithesis of art. Nevertheless, several papers have proposed methods to bridge this gap and have obtained interesting results. In fact, some art theorists have pointed out the usefulness of specific quantifiable features in the paintings. This paper presents a method for identifying painters using color profiles of skin patches in painting images. Various color models for representing the color profiles were explored. Various implementations of multiclass support vector machine classfiiers were compared. We found that a weighted combination of several directed acyclic graph SVMs with Gaussian kernels gives the best classification performance.

Despite the many significant advances made in robot motion research, few works have focused on the tight integration of high-level deliberative planning with reactive control at the lowest level. In particular, the real-time performance of existing integrated planning and control architectures is still not optimal because the reactive control capabilities have not been fully realized. This paper aims to enhance the low-level reactive capabilities of integrated planning and control with Cooperative Extended Kohonen Maps for handling complex, unpredictable environments so that the work-load of the high-level planner can be consequently eased. The enhancements include fine, smooth motion control, execution of more complex motion tasks such as overcoming unforeseen concave obstacles and traversing between closely spaced obstacles, and asynchronous execution of behaviors.

Histogram-based dissimilarity measures are extensively used for content-based image retrieval. In an earlier paper, we proposed an efficient weighted correlation dissimilarity measure for adaptive-binning color histograms. Compared to existing fixed-binning histograms and dissimilarity measures, adaptive histograms together with weighted correlation produce the best overall performance in terms of high accuracy, small number of bins, no empty bin, and efficient computation for image classification and retrieval. This paper follows up on the study of adaptive histograms by applying them to texture classification, retrieval, and clustering. Adaptive histograms are generated from the amplitude of the discrete Fourier transform of images. Extensive comparisons with well-known texture features and dissimilarity measures show that, again, adaptive histograms and weighted correlation produce good overall performance.

Despite the many significant advances made in robotics research, few works have focused on the tight integration of task planning and motion control. Most integration works involve the task planner providing discrete commands to the low-level controller, which performs kinematics and control computations to command the motor and joint actuators. This paper presents a framework of the integrated planning and control for mobile robot navigation. Unlike existing integrated approaches, it produces a sequence of checkpoints instead of a complete path at the planning level. At the motion control level, a neural network is trained to perform motor control that moves the robot from one checkpoint to the next. This method allows for a tight integration between high-level planning and low-level control, which permits real-time performance and easy modification of motion path while the robot is enroute to the goal position.

Neural networks (NNs) have been successfully applied to solve a variety of application problems including classification and function approximation. They are especially useful as function approximators because they do not require prior knowledge of the input data distribution and they have been shown to be universal approximators. In many applications, it is desirable to extract knowledge that can explain how Me problems are solved by the networks. Most existing approaches have focused on extracting symbolic rules for classification. Few methods have been devised to extract rules from trained NNs for regression. This article presents an approach for extracting rules from trained NNs for regression. Each rule in the extracted rule set corresponds to a subregion of the input space and a linear function involving the relevant input attributes of the data approximates the network output for all data samples in this subregion. Extensive experimental results on 32 benchmark data sets demonstrate the effectiveness of the proposed approach in generating accurate regression rules.

3D digitization devices produce very large sets of 3D points sampled from the surfaces of the objects being scanned. A mesh construction procedure needs to be applied to derive polygon mesh from the 3D point sets. As the 3D points derived from digitization devices based on digital imaging technologies are inherently non-uniformly distributed over regions that may contain surface discontinuities, existing methods are not suitable for polygonizing them. This paper describes a novel polygonization algorithm for constructing triangle mesh from unorganized 3D points. In contrast to existing methods, this algorithm begins the mesh construction process from 3D points lying on smooth surfaces, and advances the mesh frontier towards 3D points lying near surface discontinuities. If 3D points along the edges and at the corners are sampled, then the algorithm will form an edge where two advancing frontiers meet, and a corner where three or more frontiers meet. Otherwise, the algorithm constructs approximations of the edges and corners. It can be shown that this frontier advancing algorithm performs 2D Delaunay triangulation of 3D points lying on a plane in 3D space.

Color histogram is an important part of content-based image retrieval systems. It is a common understanding that histograms that adapt to images can represent their color distributions more efficiently than histograms with fixed binnings. However, among existing dissimilarity measures, only the Earth Mover's Distance can compare histograms with different binnings. This paper presents a detailed quantitative study of fixed and adaptive binnings and the corresponding dissimilarity measures. An efficient dissimilarity measure is proposed for comparing histograms with different binnings. Extensive test results show that adaptive binning and dissimilarity produce the best overall performance, in terms of good accuracy, small number of bins, no empty bin, and efficient computation, compared to existing fixed binning schemes and dissimilarity measures.

Texture is an important visual feature for content-based image retrieval. An ideal content-based retrieval system should compare images in its database with the query in a manner that is consistent with human's perception of visual similarity. Moreover, texture matching should be invariant to texture scale and orientation because the same texture can appear in the images in varying scales and orientations. In practice, however, texture similarity computed using computational texture features is not necessarily consistent with human's perception. This paper presents a method of mapping texture features into a texture space that is scale and orientation invariant, and at the same time, consistent with human's perception. Test results show that this method achieves a better retrieval performance than methods that are not invariant and not perceptually consistent.

An ideal retrieval system should retrieve images that satisfy the user's need, and should, therefore, measure image similarity in a manner consistent with human's perception. However, existing computational similarity measures are not perceptually consistent. This paper proposes an approach of improving retrieval performance by improving the perceptual consistency of computational similarity measures for textures based on relevance feedback judgments.

This paper describes a system for rapid acquisition of 3D models of objects for use in applications such as CAD and VR. The system consists of an image capturing platform, which comprises a PC-controlled turntable and CCD camera, and associated computer vision and graphics algorithms for 3D model acquisition. An algorithm for accurately recovering 3D points from image sequences is described. Experiments performed on synthetic and real data show that the algorithm is accurate in recovering the coordinates of 3D points and is robust against noise in 2D feature location and 3D object rotation. Results of applying the system to acquiring 3D models of real objects are illustrated.

To recognize the objects in an image and to understand the image content, a computer system has to first separate the foreground objects from the background. Image segmentation and figure-ground segregation are, therefore, essential for computer image understanding. This paper describes a system called OLAG (Object-LAyer Grouping) for image segmentation and figure-ground segregation. OLAG consists of several incremental refinement steps which use colour and other visual cues such as size and compactness for grouping the image pixels. It produces, as an end result, a set of layers each containing an object or object part. Figure and ground relationships among the objects are inferred, giving their relative depths. It is shown that interesting and useful segmentation results can be obtained from the system.

This paper describes a novel method for constructing triangle meshes from noisy and non-uniformly distributed 3D points. It consists of two steps: noise point removal uses a clustering algorithm and epipolar constraint method to identify and remove noise points from the 3D points; and a constructive polygonization algorithm interpolates cleaned 3D points to construct a triangle mesh of the object.

To process a spatiotemporal pattern, a short-term memory (STM) is needed to maintain the pattern over time. In general, a STM for storing a temporal sequence of input items should possess characteristics such as acting as a first-in-first-out buffer, storing multiple items of the same type, and encoding interval-invariant and rate-invariant representations of the input items. Some of these characteristics are lacking in existing STM models. This article presents a spatiotemporal STM that possesses all the above characteristics. It is achieved using a novel method of organizing the STM by separating the storage of spatial and temporal information.

Before symbolic rules are extracted from a trained neural network, the network is usually pruned so as to obtain more concise rules. Typical pruning algorithms require retraining the network which incurs additional cost. This paper presents FERNN, a fast method for extracting rules from trained neural networks without network re-training. Given a fully connected trained feedforward network, FERNN first identifies the relevant hidden units by computing their information gains. Next, it identifies relevant connections from the input units to the relevant hidden units by checking the magnitudes of their weights. Finally, FERNN generates rules based on the relevant hidden units and weights. Our experimental results show that the size and accuracy of the tree generated are comparable to those extracted by another method which prunes and retrains the network.

VISOR is a large connectionist system that shows how visual schemas can be learned, represented and used through mechanisms natural to neural networks. Processing in VISOR is based on cooperation, competition, and parallel bottom-up and top-down activation of schema representations. VISOR is robust against noise and variations in the inputs and parameters. It can indicate the confidence of its analysis, pay attention to important minor differences, and use context to recognize ambiguous objects. Experiments also suggest that the representation and learning are stable, and behavior is consistent with human processes such as priming, perceptual reversal and circular reaction in learning. The schema mechanisms of VISOR can serve as a starting point for building robust high-level vision systems, and perhaps for schema-based motor control and natural language processing systems as well.