A novel block matching algorithm based on particle swarm optimization with mutation operator and simplex method
Ping Zhang, Ping Wei, Hongyang Yu · 2011
To improve the performance of motion estimation in video coding, we propose a novel block matching algorithm, which utilize the global search ability of particle swarm optimization (PSO) with mutation operator and the local search ability of simplex method (SM). According to the center-biased and temporal-spatial correlation feature of motion vector and the global randomness of PSO, the fixed and random points are selected as the initial individual. Then, block matching process is executed by the updating of the position and velocity of individual. In order to accelerate the convergence of PSO and improve the accuracy of local search, mutation operator and simplex method are used. Meanwhile, based on the feature of static macro blocks, the proposed algorithm intelligently uses early termination strategies. Experimental results demonstrate that the proposed algorithm has better PSNR values than conventional fast block matching algorithms especially for video sequences with violent motion while the computational complexity of the proposed algorithm has negligible increase.