A texture segmentation method using unsupervised and supervised neural networks

Abstract

This paper deals with a segmentation method of an image composed of some kinds of textures with randomness by using unsupervised and supervised neural networks. After a texture image is divided into a number of small windows with the same size, the feature vector in those windows is extracted by using two-dimensional autoregressive model and fractal dimension. The clustering of feature vectors is performed by applying the self-organizing algorithm which is an unsupervised neural network, and the decision-based neural network which is a supervised neural network. The feature vectors which are classified by the decision-based neural network are mapped to the original image. This method has the superior segmentation ability than the method which uses both self-organization algorithm and backpropagation algorithm. >