Texture Segmentation Based on Pattern Maps Obtained by Independent Component Analysis

摘要

In this paper, we propose a new feature for texture segmentation that is based on the pixel patterns and thus is independent of the variance of illumination. A gray scale image is transformed into a pattern map in which edges and lines (bars) used to characterize the texture information are classified by pattern matching. The Gabor filters can enhance edge features, however, are not effective in edge pattern classification. We extract the pattern templates from image patches by Independent Component Analysis. Based on the pattern maps, the feature vector is comprised of the numbers of the pixels belonging to each pattern. The calculation of the features is simple and not related to the number of the components, so the proposed method is quite time saving compared with other multichannel segmentation algorithms.