Efficient Texture Classification Using Independent Component Analysis

Abstract

Texture classification is the assignment of texture to one or more texture classes. It has been largely used in various fields. This paper proposes a system for Texture Classification using Independent Component Analysis (ICA) using set of classifiers. Independent Component Analysis proved its efficiency in many domains. Our objective is to improve texture classification by adopting the use of ICA with a classifier in this domain. After extracting the main features of the image, classification using set of classifiers is performed. Experimental results have shown that ICA has a promising performance in texture classification. When combined with neural networks, Texture classification accuracy reached the accuracy of 91%. Furthermore, Naïve Bayes showed both exceptional training and testing times, and therefore, it proved to be efficient for big datasets.