A Genetic Search Algorithm for Motion Estimation

Motion estimation is essential for many interframe video coding techniques, block matching algorithms, such as FSA and TSS, have been widely used for motion estimation. The easiest implementation is the FSA, which evaluates all the blocks in the search window and has the highest computational cost. Therefore,many fast search algorithm including TSS, have been proposed to reduce the computational complexity, but most of them are based on the assumption that there should be only one optimal solution in the search window, however, in normal cases, there always exist multitudinous local optima, so they will miss the global optima, but get a suboptimal solution. In this paper, we propose a genetic search algorithm for motion estimation(GSAME) which applies genetic operation to motion estimation. We also introduce a scheme called competition evolution, which can bring the better solutions into the next evolution, and can accelerate the iteration process converging. In this method, the motion vector of block is defined as chromosome, after crossover, mutation and competition evolution, the global optimal solutions will be got. Last we compare the GASME to TSS, FSA, and the result shows that the method not only solve the problem of being trapped to local optima, but also have speed close to that of TSS.