We present an average sense optimal vector transform (AOVT) for vector quantization (VQ)-based coding of square block vectors. We introduce new notions of average intervector correlations and exploit the new structure of the intervector correlation matrix in deriving our AOVT. We present the results of applying our AOVT to many real images and demonstrate the superiority of the AOVT in terms of better preservation of the intravector correlation in the transform domain. We also present coding results on a variety of images to demonstrate the fact that the above-mentioned advantage of AOVT translates into coding gains since the performance of VQ increases if the transform domain vectors have more intravector dependencies.

This paper presents a novel multiresolution approach to the classification of textures using wavelets. The approach uses an overcomplete wavelet decomposition, called wavelet-frames, which yields the descriptions of both translation invariance and stability. In order to adapt it to the quasi-periodic properly of textures, we first detect the channels containing dominant information, and then zoom it into these frequency channels for further decomposition. For classification efficiency, we develop a progressive texture classification algorithm, in which the classification process terminates once a suitably chosen discrimination criterion is met. Experiments show that with a minimum number of wavelet frame decompositions and iterations, our proposed approach achieves a 100% correct classification rate on all the texture types tested. It outperforms many of the existing approaches in terms of classification excellence and computational efficiency, and hence appears attractive for real-time applications involving texture-based video/image classification.

We have developed a generalized alphanumeric character extraction algorithm that can efficiently and accurately locate and extract characters from complex scene images. A scene image may be complex due to the following reasons: (1) the characters are embedded in an image with other objects, such as structural bars, company logos and smears; (2) the characters may be painted or printed in any color including white, and the background color may differ only slightly from that of the characters; (3) the font, size and format of the characters may be different; and (4) the lighting may be uneven. The main contribution of this research is that it permits the quick and accurate extraction of characters in a complex scene. A coarse search technique is used to locate potential characters, and then a fine grouping technique is used to extract characters accurately. Several additional techniques in the postprocessing phase eliminate spurious as well as overlapping characters. Experimental results of segmenting characters written on cargo container surfaces show that the system is feasible under real-life constraints. The program has been installed as part of a vision system which verifies container codes on vehicles passing through the Port of Singapore.

The image correspondence problem has generally been considered the most difficult step in both stereo and temporal vision. Most existing approaches match area features or linear features extracted from an image pair. The approach described in this paper is novel in that it uses an expert system shell to develop an image correspondence knowledge-based system for the general image correspondence problem. The knowledge it uses consists of both physical properties and spatial relationships of the edges and regions in images for every edge or region matching. A computation network is used to represent this knowledge. It allows the computation of the likelihood of matching two edges or regions with logical and heuristic operators. Heuristics for determining the correspondences between image features and the problem of handling missing information will be discussed. The values of the individual matching results are used to direct the traversal and pruning of the global matching process. The problem of parallelizing the entire process will be discussed. Experimental results on real-world images show that all matching edges and regions have been identified correctly.

Fuzzy set theory is investigated as a tool for the diagnostics of systems described by means of a fault tree. The objective is to diagnose component failures from the observation of fuzzy symptoms using the information contained in a fault tree. A two-step procedure is used to solve the problem. In this first step, causal reasoning is used to diagnose failure modes, consisting of minimal cut-sets of basic events, from the observation of triggered gates treated as symptoms. In the second step, the authors identify the particular components which have failed based on the diagnosed failure modes. To perform this second step, the solution of a fuzzy relational equation a= V-product (S/sup T/ alpha x) connecting failure mode a to basic events x is derived. With this method, the diagnostics equations can be symmetrically generated and solved in terms of the tree's basic events. The systematic nature with which a diagnosis can be generated from a fault tree lends this method to potential application of object-based programming techniques.>