In the context of very low bit-rate video coding, pre-defined fixed pattern representations of moving regions in blocked-based motion estimation and compensation has become increasingly attractive over H.264 as the former represents an MB by a smaller size moving region covered by the best available pattern that approximates the shape of the region more closely and hence, requiring no extra motion vector, which is not the case with the latter. But fixed set patterns sometimes fail to code efficiently for all video sequences. In this paper a novel idea of selecting a subset of best-matched patterns through preferential selection technique is developed by presenting two algorithms. Variable Pattern Selection (VPS) and Extended VPS (EVPS) for an initial pattern codebook size of 32 using a new parametric macroblock classification definition and a new similarity metric. The complexity analysis confirmed that EVPS guaranteed to be nearly six times faster than VPS, with the peak performance providing an improvement factor of nine times. The overall performance of EVPS is identical to VPS for certain parameters but on average, 0.2dB and 0.8dB better than the contemporary algorithm using fixed set patterns and Advanced Video coding standard (H.264) respectively, for the same number of bits per frame.
|Number of pages||11|
|Journal||Journal of Research and Practice in Information Technology|
|Publication status||Published - 2005|