A single layer neural network for texture discrimination

The authors present a neural net texture classification strategy, consisting of a novel feature extraction scheme and a neural network discriminator. Texture recognition is based on the co-occurrence of n-pixel groups in the images. This results in a state vector transformation of the image. The vectors are then classified using a single layer neural network in order to identify the textures. The procedure has been tested on natural textures and the results achieved have been promising.<>