Adaptive segmentation of textured images using linear prediction and neural networks

Abstract

An adaptive technique for classifying and segmenting textured images is presented. This technique uses an efficient least squares algorithm for recursive estimation of two-dimensional autoregressive texture models and neural networks for recursive classification of the models. A network with fixed, but space-varying, interconnection weights is used to optimally select a small representative set of these models, while a network with adaptive weights is appropriately trained and used to recursively classify and segment the image. An online modification of the latter network architecture is proposed for segmenting images that comprise textures for which no prior information exists. Experimental results are given which illustrate the ability of the method to classify and segment textured images in an effective way. >