Image segmentation is the key step in object recognition and image understanding. Split-and-merge is an efficient and practical method to segment an image. The Split-and-merge method based on edge information is presented. The edge information is used both to adaptively determine the beginning layer which is optimum

and to measure the consistence of image blocks during merging and spliting. By using the method the segmentation quality is improved

the processing time is decreased

and the satisfied experiment result is obtained in segmentating some sophisticated images.