In video processing, a significant problem is the motion estimation, since it corresponds to approximately 80% of the computational effort in video coding systems. In part this complexity is due to the measure of similarity used in the block matching. In this article, instead of using a measure that evaluates every pixel within a block, it uses one that selects a subset of these pixels induced by a sequence of low discrepancy. The proposed method is applied to the Three Step Search (TSS), typically used as a technique for motion estimation in video applications video at a low transfer rate. The results show a reduction in computational complexity of about 90% with regard to the TSS, with little degradation in the recovered frame.
Motion estimation; video coding; low-discrepancy sequences