A predictive and pattern based PSO approach for motion estimation in video coding

Motion Estimation (ME) plays an important role in video compression. In Block Matching (BM) based ME approach, Exhaustive search algorithm (ES) is the initial algorithm which finds the Motion Vector (MV) through Mean Absolute Difference values for all macro blocks of the search window. But the computations required is more. Recently, several fast BM algorithms like diamond search, Three step search, Four step search are proposed to reduce the number of MAD operations by calculating only a fixed subset of search locations at the price of less accuracy. Also there are approaches based on Particle Swarm Optimization (PSO) to reduce the search points and adaptive motion estimation algorithm which reduce the computation cost based on prediction approach. In this work, a new algorithm based on the combination of PSO and adaptive motion estimation is proposed to reduce the search points, computational cost and computation time in the BM based motion estimation process and its performance is compared with Exhaustive Search (ES), Diamond Search (DS) and PSO approach for ME by measuring the PSNR values.