Fast Block-Matching Motion Estimation with Variable Thresholds andPredictive Motion Vectors

• Main Authors: Chia-Sheng Hsu, 許家盛
• Other Authors: shu-Mei Guo
• Format: Others
• Language: en_US
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/63389960278350133751

碩士 === 國立成功大學 === 資訊工程學系碩博士班 === 94 === Abstract A fast block-matching motion estimation with variable thresholds and predictive motion vectors used in the motion-compensated video compression is proposed in this paper. The proposed algorithm consists of two effective steps: 1) setting the variable thresholds to specify the stationary blocks and 2) referring the motion vectors of the neighboring blocks to efficiently determine the predictive motion vectors. The all-zero and zero-motion detection algorithms are suggested to reduce the computation of motion estimation. A robust compromise between the contribution of PSNR and the determination of stationary blocks defined in this paper has been investigated. Experimental results demonstrate more or less 50% to 80% computation load can be saved in consideration of compromise between the acceptable quality and checking points. As a result, the proposed method provides an alternative way to efficiently reduce checking points, which works on all block-matching algorithms in the literature.