Extracting Dense Features for Visual Correspondence With Graph Cuts

doi 10.1109/cvpr.2003.1211420