Smooth Energy Auto-Estimation for Graph Cuts Algorithm.
Haibing Ren, Jiali Zhao, Seok-Cheol Kee · 2005
Graph cuts is one of thebest algorithms for disparity estimation. Inits energy function, the smooth energy that is necessary for achieving the disparity map with good smoothnessandgood discontinuity varies from images to images. Based on image texture information, this paper presents a novel approach to estimate smooth energy automaticallyand precisely. To train the parameters of representation, analgorithmis proposed to auto-estimate the precise smooth energy coefficient,given the stereo images and the ground-truth disparity map. The experimentalresults show that, with the estimated smooth energy, the disparity map with less gross error could be achieved.And the disparity map obtained by hierarchical algorithm with estimated smooth energy for each level is much better than the one with same smooth energy for each level. 1