Macroblock-Based Progressive Fine Granularity Scalable Video Coding
基于宏块的渐进、精细可伸缩的视频编码

The progressive fine granularity scalable (PFGS) coding is a promising technique for streaming video applications due to its much higher coding efficiency than the FGS (Fine Granularity Scalable) coding in MPEG-4 standard. However, since the original PFGS scheme only chooses its reconstructed references as frame-based, it is very difficult to achieve a good trade-off between high coding efficiency and low drifting errors. In this paper, a flexible and effective scheme is presented to control the PFGS coding at macroblock level. Three INTER modes with distinct characteristics are first proposed for the enhancement macroblock coding. One coding mode among them provides a novel method to effectively reduce the drifing errors that may occur at low bit rates. Moreover, a decision-making mechanism based on temporal predictions is developed to choose the optimal coding mode for each enhancement macroblock, which offers a consistent improved performance for sequences with different contents and formats. Furthermore, the proposed control scheme can be easily implemented without any additional computations. The experimental results show that the macroblock-based enhancement layer coding technique technique using the three INTER modes and the proposed decision-making mechanism can effectively reduce the drifting errors at low bit rates, while providing high coding efficiency at moderate or high bit rates.