A new architecture for computationally adaptive full-search block-matching motion estimation

A new architectural enhancement for adaptive elimination of redundant computations in video motion estimation is proposed. The method differs from the existing solutions in that it dynamically adjusts the amount of computations in the full search block matching (FSBM) to the varying picture content. According to experiments, an enhanced architecture maintains the highest PSNR, while making four times less computations than the traditional FSBM design. The implementation scheme is simple, yet compact.