Texture Classification Using a Two-stage Neural Network Approach

In this a?ticle, we present a two stage neural network structure which combines the selforganizing map (SOM) and the multilayer perceptron (MLP) for the problem of texture classification. The ttxhrre features are extracted using a multichannel approach. These channels comprise of a set of Gaborfilters having different sizes, orientations and frequencies to constitute N-dimensional feature vectors. The SOM acts as a clustering inechanism to map these N-dimensional feature vectors onto a kdimensional space. This in hi171 fornis the feature space to feed into MLP for training and subsequent classification. It is shown that this mechanisnt increases the inter-class separation and decreases the intra-class distance in the feature space, hence reditces the classification compIm*ty. Also, the reduction in the dimensionality of the feature space results in reduction of the leaming time of the MLP.