An effective stereo matching algorithm with Optimal Path Cost Aggregation

This paper presents a stereo matching algorithm for obtaining dense disparity maps. Our main contribution is to introduce a new cost aggregation technique of a 3D disparity-space image data, referred to as the Optimal Path Cost Aggregation. The approach is based on the dynamic programming principle, which exactly solves one dimensional optimization problem. Furthermore, the 2D extension of the proposed technique proves an excellent approximation to the global 2D optimization problem. The effectiveness of our approach is demonstrated with several widely used synthetic and real image pairs, including ones with ground-truth value.