Segmentation of textured surfaces using mixed order statistics and neural network classifiers

The problem of textured image analysis in the presence of noise is examined from a higher-order statistical perspective. The objective is to develop analysis techniques through which robust texture characteristics are extracted and used as inputs to a neural network classifier. The approach taken involves the use of autoregressive models derived from third order cumulants. The fundamental issues of the various components of the approach are described. The neural network used in this context is a multilayer perceptron trained using conjugate gradient techniques