Computation Reduction Algorithms for Motion Estimation in Low Rate Video Coders

• Main Authors: Chin-Yun Chen, 陳慶勻
• Other Authors: Jar-Ferr Yang
• Format: Others
• Language: zh-TW
• Published: 2000
• Online Access: http://ndltd.ncl.edu.tw/handle/71228545169275303637

碩士 === 國立成功大學 === 電機工程學系 === 88 === In this thesis, computation reduction algorithms for motion estimation in low rate video coders are developed. H.263 is a low rate video coding standard which is suited for typical “head-and-shoulders” video material of video teleconferencing, and video telephony. The characteristics of such image sequences are static background and small motion of body. Many blocks after coding are all-zero blocks or zero motion blocks, so many computations are unnecessary. Computation reduction algorithms for motion estimation in allusion to the above characteristics are proposed. We jointly consider motion estimation with discrete cosine transform and quantization to detect all-zero blocks ahead of discrete cosine transform and quantization. Then several thresholding methods based on zero motion block detection are used to achieve further computation reduction. The simulation results show that much unnecessary computational effort in video coder can be saved, and make real-time video encoding possible.