Performance optimization for motion compensated 2D wavelet video compression techniques

In this paper we present two performance optimization methods for a motion compensated (MC) 2D wavelet video coding technique, which is based on two of the current state-of-the-art codecs: H.26L TML9.4 and JPEG2000 VM7.2. First, a new metric for motion vector selection is proposed to take both edge and texture complexity into account in motion prediction. Second, a frame level rate allocation algorithm, which is an extension of JPEG2000 PCRD (Post Compression Rate Distortion) optimization, is proposed. Experimental results demonstrate the significant performance improvements by these two techniques.