Constant-quality rate allocation for spectral fine granular scalable (SFGS) video coding

MPEG-4 has proposed FGS (Fine Granular Scalability) video coding technique that can adapt to Internet bandwidth variation. In this paper, we propose the Spectral FGS (SFGS), based on MPEG-4 FGS, and develop rate allocation schemes which are capable of achieving constant video quality. SFGS rearranges enhancement layer bitplane data by grouping them according to frequencies (from DC to high frequency AC components in the zigzag order) before each bit plane data are fed for FGS´s 1-D run-length coding. For each bitplane, we compute the rate-distortion (R-D) index of every spectral band and use them to determine the transmission priority, thus improving the transmission efficiency. As for the rate allocation of enhancement layer for SFGS, we accomplish it in two ways : 1) allocating bits between frames in a group of pictures by using frame features 2) allocating bits for all spectral bands in a group of pictures directly by using their R-D indexes. Experiments show that, the former can reduce average quality variation by up to 0.63 dB/frame relative to the traditional uniform method. The latter can however achieve almost constant quality in a group of frames, but need a larger buffer size and result in a longer time delay.