An adaptive flow-based dynamic search algorithm for block motion estimation

In this paper, a novel adaptive flow-based dynamic search algorithm (AFDS) is presented for fast block motion estimation. The rationale of the AFDS is to make a judicious choice of search points by fully adapting to the local characteristics of the images. To achieve this, the new search points in each iteration depend on the “flow” of the block which can be determined by the distribution of the minima computed in the previous iteration. Additionally, a dynamic jump search scheme is addressed which not only reduces the number of candidate points, but also refrains the search from being trapped into the local minima. Simulation results show that the AFDS provides viable performance with reduced computational complexity when compared with previous works. Moreover, two subblock schemes have been addressed as well for further reduction of computational complexity