Abstract:

Image segmentation based on graph cuts is a newly developing technology in recent years. But the memory overhead and time complexity of leading algorithms result in an excessive computational burden along with the increase in the size of images and spatial dimension. A fast method of motion segmentation based on graph cuts is proposed. To speed up the computation of graph cuts, the background pixels of the difference image are mapped into one or several points to participate in the creating and solving of network graphs. Through the experim〖JP2〗ents of 2D and 3D graph cuts, the fast method is applied to segment motion objects automatically in 2D image sequence and 3D spatiotemporal volume. The results of experiments show that this method reduces both the running time and the memory consumption of graph cuts while producing nearly the same segmentation result as the general graph cuts. It is showed that the method is fast and effective in motion segmentation.