EFBLA: A Two-Phase Matching Algorithm for Fast Motion Estimation

Introduction: The most popular algorithm for the VLSI implementation of motion estimation is the block-based full-search algorithm [1–3]. However, the full-search algorithm notoriously needs high computation load and large memory size. The high computational cost has become a major problem in the implementation of motion estimation. Many papers have proposed different ways to reduce the computation requirement of the full-search algorithm. Most of them target on the elimination of impossible motion vectors, such as SEA [4] and LRQ [5], and only perform motion estimation on the possible ones. Applying this philosophy, we have developed a two-phase algorithm that is suitable for implementation of motion estimation. Our goal is to decrease the number of block-matching evaluations without degrading the video quality such that the computation load can be significantly reduced. The algorithm, called the edge-matching first block-matching last algorithm (EFBLA), first employs the edge-matching procedure to determine candidate motion vectors and then performs the conventional block matching with the SAD criteria on the candidates. The edgematching procedure does not require complex computation; it only needs shift operations, quantisation, comparison and thresholding. Therefore, our algorithm requires fewer operations and lower memory size compared with the full-search algorithm. As a result of benchmarking and comparing to the full-search algorithm, EFBLA may significantly save the computation load by 93.9% while the degradation of PSNR is very little.