Winner Update on Walsh-Hadamard Domain for Fast Motion Estimation

Motion estimation (ME) plays an important role in video compression. Block-based ME has been adopted in most video compression standards due to its efficiency. In this paper, we propose a novel and fast block-based ME algorithm that is based on applying the winner-update search in the Walsh-Hadamard transform domain. Due to the energy packing capability of the Walsh-Hadamard transform, the modified winner update procedure is performed only on a small number of most representative transform coefficients. In addition, our algorithm can skip the search at some locations based on the predicted SAD of the current block. The superior performance of the proposed algorithm is demonstrated through experimental comparison with some representative motion estimation methods