A novel vector quantizer-based architecture for block-matching motion estimation

Block-matching algorithms (BMA) are commonly used in image sequence coding for motion estimation. Even though the estimation of block motion vectors by an exhaustive search is optimal, the computational complexity precludes its usage in many read-time codec applications. In this paper, we present a novel vector quantizer (VQ) based architecture for BMA offering sizeable reduction in the number of computations and also the number of motion vectors transmitted. We also compare the performance of the proposed algorithm with other commonly used fast search methods.