Proposes a model-based segmentation system that consists of a modeling suite and an image processing suite. The modeling suite incorporates knowledge about breast tissue characteristics, compression effects and the mammogram acquisition process. Given a specific segmentation task, this information is utilized by the system to invoke proper operators from the image processing suite. This paper discuses the segmentation system in the context of a specific segmentation task involving the identification of fibro-glandular tissue.

This paper deals with designing of intelligent man-machine interfaces capable of recognizing human gestures and arm/hand actions. Monitoring of operator's arm/hand motion is performed by coupling myoelectric sensors with video data acquisition. Myoelectric signals are used to simplify the problem of understanding motion. In order to retain constraint-free operator's environment, myoelectric sensing is limited to major joints (shoulder and elbow) and signals are used to estimate the location of the operator's hand in sagittal plane. Video data acquisition is used to interpret hand pose. The focus of the paper is on myoelectric data processing and showing that limited sensing of muscular activity simplifies the problem of estimating motion and provides sufficiently accurate estimates to guide video data acquisition. The problem of estimating arm motion is formulated by statistical pattern recognition based on a model that relates the EMG signal sensed at a joint to the angular displacement of the arm/hand.

Evaluation of an algorithm for finding a match for a random texture pattern in a large image database is presented. The algorithm was designed assuming that the random pattern may be subject to misregistration relative to its representation in the database and assuming that it may have missing parts. The potential applications involve authentication of legal documents, bank notes, or credit cards, where thin fibers are embedded randomly into the document medium during medium fabrication. The algorithm achieves image matching by a three-step hierarchical procedure, which starts by matching parts of fiber patterns while solving the misregistration problem and ends up by matching complete fiber patterns. Performance of the algorithm is studied both theoretically and experimentally. Theoretical analysis includes the study of the probability that two documents have the same pattern, and the probability of the algorithm establishing a wrong match, as well as the algorithm's performance in terms of processing time. Experiments involving over 250,000 trials using databases of synthetic documents, containing up to 100,000 documents, were used to confirm theoretical predictions. In addition, experiments involving a database containing real images were conducted in order to confirm that the algorithm has potential in real applications.

A parallel algorithm for halftoning of color images is proposed. The algorithm is based on the cellular neural networks (CNN). The concept of multilevel halftoning of gray-level images is extended to processing color images by trichromatic decomposition. Each of the three component images is processed separately by a multilayer CNN designed for automatic multilevel halftoning of gray-level images. The paper details the structure of the proposed network and shows the results of its simulation.

This paper describes designing and testing systems for inspection of web materials using a development environment simulated in software. This environment comprises five sub-systems: sensing, detection, characterization, validation and decision making. Given a specific inspection problem the system is customized by incorporating specific algorithms into each of the sub-systems. In order to illustrate the customization, the paper uses an example of inspection of a nonwoven material. The inspection objective is to quantify the randomness and homogeneity across the material. This task requires using statistical modeling to characterize texture. The paper discusses a theoretical approach based on a Poissonian model, its efficient implementation, and steps required to implement this procedure online.

This paper describes part of a study aimed at developing a computer-based aid for mammogram screening that makes a detailed comparison between mammograms of the same patient acquired at different screenings and detects changes indicative of cancer. The focus is on determining control points in two mammograms; these points are used to put two mammograms into correspondence. The paper details the algorithm for identifying the potential control points and establishing the correspondence between the two sets of control points. The algorithm's performance was evaluated by three observers, one of whom is an experienced radiologist, and found to be adequate.

The authors propose a new methodology for identification of feature points in non-structured texture images and suggest how they can be used to perform regional registration when precise (pixel-wise) registration is not possible. The specific motivation behind this work is the analysis of sequences of mammographic images of the same patient with the objective of detecting changes indicative of cancer. Because of intractability of precise mammogram registration the authors propose regional mammogram registration and comparison by using feature points. This paper focuses on extraction and pairing of the feature points. The proposed technique is tested on 50 mammogram pairs and verified by visual inspection involving three human subjects, an experienced radiologist and two casual observers.

This paper describes the development environment for the design and testing of an expert system for web material inspection. The environment incorporates five sub-systems: sensing, detection, characterization, feature analysis, and classification. Each sub-system consists of a number of modules targeted at performing specific functions and containing adaptive image processing and pattern recognition algorithms. An example of defect classification in web materials is considered. The results are summarized and interpreted.

An approach to authenticating/recognizing documents based on random patterns embedded into the document medium is described. The random patterns are formed by embedding short fibers during medium fabrication and the resulting documents have extremely high uniqueness. Therefore, these patterns offer an effective anti-counterfeiting measure. The proposed algorithm compares a document under consideration to all documents contained in a database assuming that the newly acquired document may be rotated and translated relative to its match in the database, and/or parts of the document may be missing. The algorithm has been tested on over 2000 synthetic images always giving satisfactory results.>

The authors describe a vision system that can monitor a weld pool in multipass arc-tungsten-gas welding. The system uses sequences of images acquired by a coaxial viewing where the optical axis coincides with the axis of the welding electrode. First, the system detects the weld pool edge by transforming an image such a way that a weld pool edge maps into a vertical curve that can easily be detected. Next, a weld pool is characterized by a pattern vector containing measurements pertaining to weld pool shape, size, and symmetry. The measurements were performed by using the detected weld pool edge. An analysis of a limited number of pattern vectors corresponding to normal and abnormal welding conditions indicates that such measurements have potential to be used in real-time weld pool monitoring and welding control.>

A vision system that determines the edges of the weld pool in sequences of gas-tungsten-arc welding images acquired by a coaxial viewing system is described. The vision system uses a transformation that maps the edge of a weld pool into a vertical line. The weld pool edge is detected in the transform domain by using a directional filter, which retains only intensity changes of interest, and a one-dimensional edge detector. The edge of the weld pool, in the physical domain, is determined using the inverse transformation. The transformation uses parameters that are updated when processing a sequence of images, and are initially determined by analyzing the first image frame in the physical domain.>

The vision system described reconstructs 3D scenes from sequences of noisy binocular images. First, the system establishes all possible matches between the feature pixels in the first binocular image pair and assigns a confidence value to a possible match. Each possible match is associated with a 3D volume. The possible matches are used to predict projections of associated 3D volumes onto the remaining binocular image pairs. These projections are utilized to limit searches for possible matches. The matches are used to update confidence values using an optimal fusion algorithm. Finally, matched pixels with high confidence values are used to reconstruct a 3D scene.>

A distributed algorithm for multisensor integration that considers sensors represented in local coordinate systems and the communication networks between sensors with uncertainties is described. The algorithm is applied to the problem of 3D scene reconstruction from binocular image sequences. The essence of the approach to 3D scene reconstruction is the establishment of correspondences between feature pixels in each image pair and the assignment of a 3D region to each established correspondence. The feature pixels are zero-crossings, and correspondences are established pairwise and between the frames of each sequence under specified criteria. The obtained 3D regions are utilized as the means of information integration to reduce spatial uncertainty. The algorithm has been applied to various binocular image sequences, and the results are generally very good.>

The authors describe a method for restoring sequences of noisy images obtained by acquiring different views of the same scene. The method uses a 3-D Markov random field and a least-square-error matching to establish the temporal-spatial neighborhood of a pixel in an image under restoration. The problem of image sequence restoration is posed as the problem of maximizing the conditional probabilities. This task is accomplished by a modified version of the iterated conditional modes method where Gibbs distribution is used to model the prior probability.>

The authors describe a novel algorithm for the detection of road edges in image sequences for mobile robot navigation. The essence of this algorithm is road modeling by spline functions and edge detection by a linear recursive filter. The filter uses a measurement procedure that involves an edge point search about a specified reference curve. This curve is determined automatically by preprocessing the first frame. Estimates of the road edges in each frame are used as the reference curve for the following frame. The algorithm processes a sequence of road images, and it does not require a priori knowledge about the image contents. The algorithm has been tested on image sequences, yielding satisfactory results.>

The location of a scene element can be determined from the disparity of two of its depicted entities (each in a different image). Prior to establishing disparity, however, the correspondence problem must be solved. It is shown that for the axial-motion stereo camera model the probability of determining unambiguous correspondence assignments is significantly greater than that for other stereo camera models. However, the mere geometry of the stereo camera system does not provide sufficient information for uniquely identifying correct correspondences. Therefore, additional constraints derived from justifiable assumptions about the scene domain and from the scene radiance model are utilized to reduce the number of potential matches. The measure for establishing the correct correspondence is shown to be a function of the geometrical constraints, scene constraints, and scene radiance model.>