We propose a new image segmentation method that can extract multiple gray objects from a gray image even under nonuniform illumination. Because the minimum area and boundary intensity of the object are the only two parameters used, this new method can be easily applied. The new method is based on the fact that one object can exist in a certain threshold range. Thus, if the original gray image is binarized by successive thresholds, the contours of the object may be different, but it can be found that these contours usually pass through boundary pixels more times than other pixels. We call this observation relation stable state. By using an image to record contours and a new merge-split processing to segment the image, the new method integrates gray value, edge information, and space connectivity smartly. Experiments show it can be used not only in normal cases but also in extracting multiple gray objects even under nonuniform illumination.