Fast Computation of Large Disparity Matching Approach Based on Correlation Window

摘要

In order to get the dense disparity in stereo matching, the algorithms based on correlation is applied widely, but the large computation cost is the main problem in realtime application, especially in large disparity stereopair matching. In this paper, a fast computation approach based on correlation window matching is proposed suitable to large disparity stereo matching. Firstly, the proposed approach estimates the disparity of image level and gets the new left and right images. Then, matching algorithm based on correlation window is applied to get the new disparity of the new stereopair, the computation optimization strategy is adapted to accelerate the algorithm running speed, the sum of estimation disparity and the disparity obtained in new stereo pair is the result disparity. Finally, the compared experiments are presented the paper. The experiment results indicate that proposed approach could save the running time in prodigious degree.