A variable pattern selection algorithm with improved pattern selection technique for low bit-rate video-coding focusing on moving objects

Research into pattern representation of moving regions in blocked-based motion estimation and compensation in video sequences has focused mainly upon using a fixed number of regular shaped patterns. Recently we presented the Variable Pattern Selection (VPS) algorithm, which selects a preset number of best-matched patterns from a pattern codebook of regular shaped patterns. The pattern elimination technique in the selection process of the VPS algorithm however, was to be computationally expensive, especially when the preset number is low. In this paper, the concept is extended to develop the Extended Variable Pattern Selection (EVPS) algorithm where the pattern elimination technique is replaced with a fast solution. The complexity analysis confirms that this algorithm can be as much as 8.5 times faster than the VPS algorithm. In order to take advantage of this computational speed-up in eliminating patterns, the pattern codebook size of the EVPS algorithm has also been increased to 32.