Probability halftone image model in a problem of unsupervised pattern recognition absed on directed enumeration method

The problem of automatic image recognition with unsupervised learning is put and solved by the direct enumeration method. The new probability-theoretic model of halftone image is investigated. Its application has allowed synthesizing new recognition criterion with self-training on the basis of optimum Bayesian statistical rule. The experimental results in a problem of face recognition are presented. It is shown that the proposed image model combining with directed enumeration method is characterized by frequentative calculations' reduction with preservation of image recognition accuracy.