Traditional image segmentation methods often have some disadvantages, such as the over segmentation, the lower segmentation speed etc. In order to solve the above problems, a new region-based color image segmentation method is proposed in the paper, which transforms an image to L*a*b* space, and divides the image into sub-blocks, and extracts the color, texture and location features for every sub-block with combining the harr wavelet transform firstly. Then, the class number and the initial clustering centers are produced by using an improved Mean-shift algorithm. Finally, the method adopts an...