Mean shift is a widely used clustering algorithm in image segmentation. However, the segmenting results are not so good as expected when dealing with the texture surface due to the influence of the textures. Therefore, an approach based on wavelet transform (WT), co-occurrence matrix (COM) and mean shift is proposed in this paper. First, WT and COM are employed to extract the optimal resolution approximation of the original image as feature image. Then, mean shift is successfully used to obtain better detection results. Finally, experiments are done to show this approach is effective. Mean, is a widely used clustering algorithm in image segmentation. Thus, an approach based on wavelet texture (WT), co First, WT and COM are employed to extract the optimal resolution approximation of the original image as feature image. Then, mean shift is successfully used to obtain better detection results. Finally , experiments are done to show this approach is effective.