Content-based motion estimation with extended temporal-spatial analysis

In this paper, a content-based motion estimation algorithm that extends the utilization of spatial correlation into the reference frame is proposed. This algorithm incorporates adaptive search window (ASW) and three step search (TSS), and realizes dynamic search strategy optimization according to local motion complexity of video contents with a new motion type prediction technique that examines local motion vector distribution of two consecutive frames. Simulation results show that the proposed method can constantly reduce the dynamic computational cost to as low as 3%∼4% of that of full search (FS), with the visual quality in terms of PSNR remaining very close to that of FS for various test video sequences.