An adaptive fast full search motion estimation algorithm for H.264

In this study, an adaptive fast full search (FFS) motion estimation (ME) algorithm for H.264 is proposed. In the proposed algorithm, first, a proposed modified data reusing (MDR) technique is used to speed up the ME processing time by reducing sum of absolute differences (SAD) computations. in the proposed rood search order (RSO), the search order can be determined in such a way that the distribution of MV/spl I.bar/COST values is "linearly-related" with the search order, i.e., MV/spl I.bar/COST is a monotonically nondecreasing function of the search order so that the number of MV/spl I.bar/COST computations can be greatly reduced. Second, the proposed adaptive search order (ASO) technique is used to speed up the ME processing time by reducing MV/spl I.bar/COST computations and adaptively switching the ME search order between the spiral search order and the proposed rood search order (RSO). Based on the simulation results obtained in this study, the performance (the average processing time) of the proposed mixed algorithm (MDR+ASO) is better than that of five comparison FFS algorithms.