Abstract: The automatic seeded process expends much time in the region growing method, Focused on this problem, an image segmentation method based on sub-region growing is proposed, to advance the efficiency of image segmentation. Segmentation and automatic seeded process are combined based on the sub-regions which grow coherently, and the segmentation speed is increased. Aimed at that, image noise affects detection of region coherence, the mean distance of region pixels used as the standard of region coherence is presented at the same time. Experiments showed that the mean distance of region pixels is better than traditional squares of subtraction on the capability of noise suppressed; Experiments of visual image segmentation of mobile robot showed the presented method to be effective.