Fast disparity estimation algorithm for mesh-based stereo image/video compression with two-stage hybrid approach

摘要

Disparity estimation is a very important operation in stereo image and video compression. However, existing disparity estimation algorithms require too large computation power. In this paper, we propose a fast disparity estimation algorithm for mesh-based stereo image and video compression with two-stage hybrid approach, which is names as Two Stage Iterative Block and Octagonal Matching algorithm (TS-IBOM). The first stage of this algorithm is Iterative Block Matching algorithm, which can give a good initial guess of the disparity vectors with a little computation. In the second stage, Iterative Octagonal Matching algorithm is employed to refine the disparity vectors. A local updating scheme is also proposed to further accelerate the process by skipping nodes whose disparity vectors have been derived already. Experimental results show that the proposed algorithm can give good results and is 18 times faster than Octagonal Matching. This algorithm is suitable to be employed in mesh-based stereo image and video compression systems.