Object segmentation based on watershed and graph cut

In this paper we select the inner contour and outer contour of the target object, and then apply watershed algorithm to divide the region between the inner contour and the outer contour into series of smaller regions. Each smaller region is regarded as a node, and the small regions contains of the inner contour are regarded as the source s, and the small regions contains of the outer contour are regarded as the sink t. Then from the source to the sink establishes a graph. Merging the capacities of parallel edges, deleting unnecessarily edges, multi-source and multi-sink can be converted to the single source and the single sink. And then minimum energy can be obtained by the maximal-flow cut, and the target object can be extracted. The experiment proves that this algorithm is effective in the segmentation than the other traditional algorithms, and has better performance in the running time.