摘要
精细的可伸缩性的视频编码 FGS(Fine Granular Scalable)是 MPEG- 4标准的视频流化框架中的关键性编码技术 ,由于在 FGS编码方案中运动补偿是参考一个最低质量的重构层 ,因而编码效率较低 .在这篇文章中 ,我们首先提出了一个渐进的精细可伸缩性的视频编码框架 ,简称为 PFGS(Progressive Fine Granular Scalable) ,与MEPG- 4中的 FGS相比 ,PFGS编码框架试图在增强层的编码过程中采用多个高质量的参考来提高编码效率 ,这是因为高质量的参考可使运动预测更准确 .但是 ,高质量的参考也会带来一些问题 .首先 ,增加多个参考会使得PFGS编码需要更多的内存 ,同时也会增加其计算复杂性 ,因而在这篇文章中 ,首先将 PFGS编码框架简化为只需采用一个额外的缓存来存储参考图像 ,而得到的编码效率几乎和原来多个参考一样 .其次 ,在编码过程中 ,由于转换参考图像会造成增强层间编码系数的振荡 ,从而部分地抵消了采取高质量参考的优势 .文中同时提出了一个有效的消除振荡的方法 ,即采用高质量的参考对所有增强层进行编码而用低质量的参考来重构以解决误差传播问题 .实验结果表明 ,我们所提出的 PFGS编码框架的编码效率能比 MPGE- 4中的 FGS提高 1d B以上 ,同时保留了FGS所有的优点 ,比如精细可调性 ,自适应网络带宽变化?
Fine Granularity Scalability (FGS) video coding is a key technique in MPEG 4 streaming video profile. However, since it always uses low quality reconstructed images as reference frames in motion compensation, it has relatively low coding efficiency. This paper first proposes a basic framework, Progressive Fine Granular Scalable (PFGS), which provides a highly efficient scalable video coding. Compared to MPEG 4, the PFGS framework uses multiple references with increasing quality for coding efficiency because high quality references make the motion prediction more accurate. However, using multiple different quality references also introduces several issues. Firstly, multiple extra frame buffers are needed for saving the reconstructed layers. This increases the memory cost and computational complexity of the PFGS system. Based on the basic framework, a simplified and efficient PFGS framework is further proposed. The simplified PFGS framework needs only one extra frame buffer with almost the same coding efficiency as the original framework. Secondly, there are undesirable increases and fluctuations in the residue coefficients during switching references form a low quality to a high quality, which partly offset the advantage of high quality references. Furthermore, a scheme eliminates the fluctuations among enhancement layers by using the high quality reference for all enhancement layers coding and solving the error drifting problem through reconstruction using low quality references. Experimental results show the PFGS framework can improve coding efficiency up to more than 1dB compared with Fine Granular Scalability (FGS) on the average PSNR, yet still keeps all the original properties, such as fine granularity, bandwidth adaptation and error recovery.
出处
《计算机学报》
EI
CSCD
北大核心
2000年第12期1276-1282,共7页
Chinese Journal of Computers
