Efficient coding of motion vector and mode information for H.264/AVC

摘要

This paper aims to reduce the size of header in H.264; therefore we only focus on compressing the larger header data into small one. H.264/AVC is the newest video coding standard, which adopts some new techniques to improve the compression efficiency. However, H.264 gets larger header than those of previous standards because the techniques exploit in detail the content of input image. The most parts of header are composed of macroblock type (mb_type) and motion vector difference (mvd), and there are redundancies in the header of H.264. Especially, when mode 8 is selected during mode decision at macroblock level, all four sub-macroblock data are transferred into bitstream. Building the idea above, we propose the scheme using the type code and quadtree to solve the problems. Experimental results show that our algorithm achieved lower total number of encoded bits over JM14.2 up to 31.87% bit reduction at QP 40 on average.