Hierarchical disparity estimation using spatial correlation

Abstract

In this paper, we propose a fast hierarchical algorithm for disparity estimation, which is based on the multiresolution block matching technique with prospective disparity vectors. The disparity vectors of each template block's spatially adjacent blocks are adopted as the initial references for the template block to speed up the estimation process. The experimental results show that the proposed hierarchical estimation algorithm can reduce computation cost drastically while maintaining comparable PSNR quality. The proposed scheme also results in smoother disparity vector field compared to the exhaustive search scheme. A rate-distortion optimized hierarchical disparity estimation algorithm is also proposed to minimize the distortion of the reconstructed image sequence under a target bit rate constraint. The experimental results show that its PSNR quality is better than the full search algorithm while taking more computing power for the optimization process.