Abstract: This study proposed two algorithms, an adaptive motion estimation algorithm based on Bayesian decision and an improved Motion Vector Field Adaptive Search Technique (MVFAST) algorithm. The first algorithm proposed in this paper, firstly, statistic the current frame and the before frame coded macro block information, Then, using the Bayesian theory, combined with the consistency characteristics of macro block MV space, to get an accurate and fast prediction motion vector. The second algorithm proposed in this paper with better early termination strategy, that is to set up a dynamic threshold of the threshold and take full advantage of the spatial and temporal correlation of video sequences, the use of the adjacent macro block motion vector to block movement by type starting point for prediction using different search strategies on the macro block.