A new stereo matching method with combination of cross-based aggregation and hierarchical belief propagation

This paper proposed a new stereo matching method with a combination of cross-based support aggregation based on the assumption that neighboring pixels with similar colors should have the same disparities and an efficient belief propagation optimization algorithm. With the merits of both local method and global method, we first propose a novel matching cost function incorporating the edge information to reduce disparity errors around edges. Then, a cross-based algorithm is utilized for aggregation process, followed by an optimization computing adopting hierarchical belief propagation algorithm. We also do some disparity refinements including right-left consistency checking and sub-pixel enhancement. Experimental results show that our method is effective and reasonable.