Texture feature extraction using ICA filters

Abstract

A novel image texture extraction approach using Independent Component Analysis (ICA) filters for image classification is proposed in this paper. Firstly groups of filters (ICA filters) are extracted from the sample texture images using the ICA method. And then, ICA filters are evaluated and selected according to the response of the input sample images to these filters for the purpose of reducing feature dimension. Finally, global and local features are extracted from the histogram of the maximum response of the input test image to the selected filters. Experimental results show that the proposed texture feature has better classification correct rate than that of MPEG-7 texture descriptors.