Abstract: An adaptive multi-pattern refinement algorithm for motion vector in video transcoding is presented.Using the consistency of angle and amplitude of intensity of local motion vectors,a motion activity model is constructed,so the search range of the macroblock is confirmed adaptively using the activity model.Then,a multi-pattern refinement strategy based on the diamond search and horizontal and vertical search is proposed.Experimental results show that the proposed method reduces the computation complexity significantly without degrading the quality of video transcoder.