Video Image Block-matching Motion Estimation Algorithm Based on Two-step Search
Wei-qi JIN,Yan CHEN,Bin LIU,Chong-liang LIU,Ya-zhong SHEN,Gui-qing ZHANG
2010-01-01
Abstract:Aiming at the shortcoming that certain existing blocking-matching algorithms,such as full search,three-step search,and diamond search algorithms,usually can not keep a good balance between high accuracy and low computational complexity,a block-matching motion estimation algorithm based on two-step search is proposed in this paper.According to the fact that the gray values of adjacent pixels will not vary fast,the algorithm employs an interlaced search pattern in the search window to estimate the motion vector of the object-block.Simulation and actual experiments demonstrate that the proposed algorithm greatly outperforms the well-known three-step search and diamond search algorithms,no matter the motion vector is large or small.Compared with the full search algorithm,the proposed one achieves similar performance but requires much less computation,therefore,the algorithm is well qualified for real-time video image processing.