Abstract: In order to improve the quality of image segmentation, graph theory and minimal cut set algorithm were used. Firstly, using the pixel points of image as the mapping nodes of the graph theory, the node weight were calculated by the ratio of the balance factor and the shared nearest neighbor nodes. Then, the minimum cut set of the image was established based on the minimized energy equation, the gray value of the segmentation block was extracted as the block feature vector and the image was segmented by minimum spanning tree. The adjacent regions were judged to be combined or to be segmented by judging function. Finally the algorithm flow was given. The results show that the target information can be segmented by this algorithm. This algorithm has good robustness and small peak memory.