Fast and Adaptive Motion Vector Search

The unacceptable computational cost of motion estimation(ME) using full search(FS) led to vast research in the field.A fast and adaptive ME algorithm is presented in this paper.The correlation of spatial and temporal neighboring blocks is used firstly to predict an initial motion vector(MV) for the current block.By the region motion intensity,various search patterns are chosen adaptively to refine the MV.To avoid the unnecessary search,the halfway stop technique is adopted.Combined with the check priority rule,the search is sped up further.Experimental results show the proposed algorithm provides the competitive performance with lower computational complexity.