Classification of Random Image Fields Using Synthetic Discriminant Functions: Spectral Statistical Approach and Its Computer-Optical Realization

The problem of classification of images that have a perfectly random nature is considered. We propose a new approach to solve this problem that is based on the use of the synthetic discriminant functions being synthesized to separate linearly the power spectra of random image fields to be classified. In the stage of both discriminant function synthesis and classification, the statistical technique of power spectrum estimation is employed. The realization of the proposed approach by means of a hybrid computer-optical technique is discussed, and its efficiency is illustrated by two examples of real-world texture classification.