FEATURE-EXTRACTION FOR TEXTURE-DISCRIMINATION VIA RANDOM-FIELD MODELSWITH RANDOM SPATIAL INTERACTION

In this paper, we attack the problem of distinguishing textured images of real surfaces using small samples. We first analyze experimental d ata that results from applying ordinary conditional Markov fields, In the face of the disappointing performance of these models, we introduc e a random field with spatial interaction that is itself a random vari able (usually referred to as a random field in a random environment), For this class of models, we establish the power spectrum and the auto correlation function as well-defined quantities, and we devise a schem e for the estimation of related parameters, The new set of features th at resulted from this approach was applied to real images. Accurate di scrimination was observed even for boxes of size 16 x 16.