Research On Color Image Segmentation Algorithm

Image segmentation algorithm is one of the important steps in image processing. It will directly impact on the following study. In the past, most efforts have been exerted in monochrome image segmentation algorithm. With the rapid development of science and technology and the related hardware devices, color image processing techniques have obtained more and more attention. At present, most of the color image segmentation algotithms are based on monochrome image. They always ignore the inherent relation present in three color components when studing color image processing. It will cause poor result.Firstly, the color spaces which have appeared in color image segmentation field have been analysed and summarized, and then we concluded the characteristics of them. Some major color image segmentation techniques have been analysed, classified and summarized. And we compared the advantages and disadvantags of them.Secondly, to maintain inherent relation of three color components, we use quaternion model for color image to solve. To improve the efficiency of color image segmentation algorithm, we use quaternion principal component analysis to process color images. Due to study th basic theory of QPCA, the general steps of quaternion principal component analysis have been summarized. Based on the analysis and experiments, we concluded approach for color images. We propose a new algorithm which is based on QPCA and algebraic multi-grid method. Our approach is to extend the monochrome image segmentation algorithm which is based on multi-grid to color image technology. And we achieve this method in quaternion domain.Thirdly, we do some experiments for our algorithm, and compare with three mainstream methods. The results demonstrate that our algorithm is effective.In the very end, based on a summary of the research results, several questions for future work and the further development of image segmentation are proposed.