A new digital image segmentation method based on regularization on graphs is proposed, which applied a regularized diffusion framework to solve the image segmentation problem with supervised learning. The weight of the graph is generated by using Gaussian Kernel Function, combining with the geometric feature extracted from the image with contourlet transform and the color feature with HSI decomposition. The graph topology structure is an improved 8-connection topology whose step is 2k,k=0,1,2,3. Experimental results show that, compared with some graph spectral theory based image segmentation algorithms, such as Random Walker and the Lazy Snapping, the proposed method is robust for noisy pictures, which can reserve more complete boundary and have better performance on the section with inconsistent texture.
