DSRA - A BLOCK MATCHING ALGORITHM FOR NEAR-REAL-TIME VIDEO ENCODING

The correlations of motion vectors between neighboring image blocks ha ve long been applied to improve the efficiency of block matching algor ithms. In this paper, a block matching algorithm with a dynamic search range is presented. The search range is adjusted based on the correla tions of the motion vectors between neighboring blocks. It is shown by simulation results that the proposed method can speed an MPEG-1 encod ing system up to a ratio of 13.3 and increases the bit-rate only of 0. 025 bpp, in average. It is therefore beneficial to embed the proposed method into a motion-based encoding system.