Abstract
As can be observed from the literature survey, there is no commonly accepted quantitative definition of visual texture. As a consequence, researchers seeking a quantitative texture measure have been forced to search intuitively for texture features, and then attempt to evaluate their performance by different techniques. Dissimilarity analysis is one of the main requirements from the classifier design point of view and provides information of significant importance - regarding feature extraction and selection strategies. This paper explores several texture features of historical and practical significance and presents their comprehensive dissimilarity analysis. An improved post processing scheme has also been proposed for Law's filter based feature extraction technique. Results show a substantial improvement over existing scheme. Cross validation of the results has been accomplished through supervised classification using Probabilistic Neural Network.