高动态范围视频的量化编码技术研究

The high dynamic range (HDR) of the luminance has become the standard indicator of the new generation of the digital television system. Meanwhile, the HDR video will also present more realistic scene information to the audience. Compared to the traditional video, the HDR video has recorded the information of higher luminance range and finer contrast. However, the current video processing system has been designed based on the data characteristics of the traditional low dynamic range video, which supports the low contrast accuracy and the limited luminance range. Therefore, the research on the reasonable degradation and data representation of the HDR video information is a key scientific issue for the application of the HDR video in current video processing systems. In this research, the applicant proposes a new method for the quantization of the HDR luminance based on the contrast threshold and the image saliency in order to better preserve the contrast information. Meanwhile, with this quantization method, a new coding solution for the HDR video is proposed. This solution is designed to preserve the HDR video information more efficiently than the one of the standard. It is also expected to be helpful for the development of the High Definition video communication in China.

高动态亮度范围已经成为新一代数字电视系统的标准指标，高动态视频也将为人们呈现更加逼真的场景信息。相对于传统视频，高动态视频记录的亮度范围更广，对比度更为精细。但是当前视频处理系统的设计主要考虑了传统低动态视频的数据特性：对比度精度低，同时支持的亮度范围有限。因此研究高动态视频信息的合理退化与数据表示，是将高动态视频信息应用在当前视频处理系统的关键科学问题。在本项研究中，申请人以人类视觉的对比度门限和图像显著性为基础，提出新的高动态亮度信息量化方法，以更完善地保留对比度信息。同时，基于这种量化方法，提出新的针对高动态视频的编码方案。该方案力争比国际标准建议的方案更有效地保存高动态视频信息，并期望为我国高清视频通信的发展提供帮助。

高动态视频可以展示低动态视频不具备的对比度细节和色彩，当前高动态视频的应用正在学界和产业界引起极大关注。本研究针对高动态视频的应用课题，展开对高动态图像视频量化和编码方法的研究工作。基于人类视觉特性，本研究在高动态量化工作中，将高动态亮度范围划分成若干亮度区间，并设计显著性因子和分布因子用于亮度区间的量化权重计算。亮度区间的量化权重值可以决定这个亮度区间的量化精度，进而实现针对高动态视频的自适应量化算法。在量化算法的基础上，本研究设计并实现了高动态视频编码方案，同时，引入了防闪动功能，提高视频视觉质量，为高动态视频的存储和传输提供了参考。实验结果证明本研究设计并完成的高动态视频编码方案，相对于近年来其它高动态视频编码方案，在编码所需的比特率和解码后视频的质量上，都有很大的优势。具体表现为，在达到特定解码视频质量时，本研究的方案所需的编码比特率更少；在所需同等编码比特率条件下，本研究方案的解码高动态视频质量更高。