The Study Of Color Image Segmentation Method Based On The Characteristics Of HVS

Image segmentation is the basic task for image recognition and interpreting. Its applications includes computer vision of mobile robots, extraction of moving objects in intelligent video monitoring systems, content based image retrieval, content based image compression, and so on. Color image has been becoming increasingly popular in recent years. Although much research has been carried out on grey image segmentation, the study on color image segmentation is still in the developing situation. Focusing on the problems in the current color image segmentation, we investigate the effective color image segmentation method based on the characteristics of human vision system (HVS). The following research has been accomplished.(1) The basic theory and methods associated with image segmentation are first analyzed and investigated, which includes the analysis of various color space systems, the extraction methods of texture feature of image, and the basic image segmentation algorithms.(2) In order to simulate the color image segmentation mechanism of human- being, this paper presents a color image segmentation method by combination of clustering and region growing techniques. To capture the texture feature of color image, the original image is first partitioned into 16×16 sub-blocks which are not overlapped, and then the sub-blocks are segmented by color clustering algorithm based on perceptual uniformity. Color and texture features are extracted from the segmented sub-blocks, and a block growing algorithm is employed to achieve segmentation for the whole image based on growing rules determined in term of human vision characteristics. The proposed method combines the advantages of clustering and region growing approaches, which is in accord with the human segmentation strategy. The new algorithm is applied to segment a number of natural images and its effectiveness and efficiency is confirmed by experimental results. Reasonable texture feature of color image is critical to color image segmentation.(3) Based on the color image segmentation method by combination of clustering and region growing techniques, the texture feature of sub-block image in DWT domain is extracted for color image segmentation in this paper, for the coefficients s of wavelet transform provide the directional characteristics and energy compacting feature. Experiments were conducted to evaluate its performance.