Object auto-segmentation based on watershed and graph cut

In order to segment CT slices mostly and accurately, watershed algorithm and graph cut were combined. Firstly, the inner contour and the outer contour of the target object were selected, and then watershed algorithm was applied to divide the region between contours into series of smaller regions. Each smaller region was regarded as a node to establish the graph. Multiple sources and multiple sinks can be converted to the single source and the single sink to refine the graph. Secondly, the target object of the first CT can be extracted by the maximal-flow cut. Thirdly, mapping the contour to the next CT, the contour was reduced and expanded to regard as the inner contours and the outer contours. Then the next CT was segmented. Followed by the cycling, the entire sequence of CT will be operated. The experiment proves that this algorithm is more effective in the segmentation and the running time than the other traditional algorithms. Meanwhile, the contours of the series of CT were extracted automatically.