A system has been developed at the University of Connecticut Image Processing Lab for the manage ment of pictorial databases. The technique is geared toward minimizing manual intervention in entering and accessing picture files and, by assign ing a loose ordering to the files, permits fast searching and clustering. The two major concepts involved are the classification of grey level histo grams to be used as the similarity measure for the ordering and a pyramid structure to obtain several levels of image classification in a hierarchical manner. The method has been tested with several sets of picture files and compared as far as ease and speed of operation against one of the more common methods currently in use with extremely favorable results.

This Special Issue is composed of the papers selected from the 1978 IEEE Computer Society Workshop on Pattern Recognition (PR) and Artificial Intelligence (Al) held in Princeton, NJ, April 12-14, 1978. The Workshop was sponsored by the Technical Committee on Machine Intelligence and Pattern Analysis. Inevitably, the contributors to the Workshop determined, to a large degree, the tone and complexion of this Special Issue. For this reason, a brief account of the Workshop Proceedings, though now history, is given. About half of the papers presented at the Workshop were also submitted for the Special Issue, a total of 37. Those of high quality were far more than the number that could be accommodated within the available number of pages. We decided to choose three topics where the interaction between the methodologies of PR and Al was most prevelant: analysis of images, analysis of speech, and certain general algorithms. All the selected papers present either theoretical, or experimental results, or both. We felt that such results clearly demonstrate the progress achieved and can be seen as very impressive if measured against the difficult problem of emulating functions associated with human intelligence by machines. It is true that they often fall short from some of the earlier ambitious goals, but the time is probably ripe to reexamine such goals in view of the accumulated experience. The following is a brief scanning of the contents of this issue, especially as related to the integration and/or interaction of PR and Al methodologies.

Simple recursive methods for inverting an n x n matrix A + E in terms of A-1and E are presented, where E represents a matrix of modifications of the matrix A. Algorithms for rank one and rank r matrix modifications are given. In addition, simple methods for determining if A + E is invertible are developed. Applications of these methods to pattern recognition problems where the inversion of a matrix (eg., covariance matrix, scatter matrix, etc.) must. be computed and frequently updated as changes in data occur are illustrated.

For a given set of images (or pattern samples), a probabilistic grammar may be developed to describe and recognize these images. This grammar, once obtained, may serve as an abstract representation for the images in question. Conversely, images may be regenerated from their grammatical representations. Images regenerated in this manner, however, may contain samples that are not found in the set of original images from which the grammatical representation was developed. A method for evaluating the quality of regenerated images with respect to the set of original images is discussed. Let I be the set of original images and O be the set of images regenerated from the abstract representation of I. Set O consists of two mutually exclusive subsets, A and B. A is the set of images found in I, and B is the set of images not found in I. It is shown that the clustering behavior of set B determines the quality of the regenerated images in the mean-square sense. By defining a scatter matrix for B, several factors that influence the image quality with respect to I are derived. Experimental results on the abstract representation and regeneration of a subset of the Munson data are demonstrated. An interesting application of this regeneration model to pattern recognition in storage-restricted environments is also presented.

A projection of a digital picture (image) may be defined as an ordered set of values, each of which is the sum of the gray levels along a given path drawn through the picture. A projection set is at ordered set of projections of the picture. Given a projection set, reconstruction is achieved by operating on the projection set to produce either the original picture or an approximation to it.