超高清全色度屏幕图像的编解码SoC系统架构与关键技术研究

The onset of cloud computing, cloud rendering and cloud-mobile computing has brought many new challenges to the video coding community. In particular, high quality, low bitrate and low complexity screen content coding is indispensable in many rapidly growing cloud and mobile related applications such as virtualized screen, thin client-server platform, remote desktop, and wireless mirror (remote or local) display for mobile devices at office or home..The image content of a screen can be classified into two types: mostly computer- generated discontinuous tone content such as text, chart, graphics and icon, and mostly camera-captured continuous tone content such as photograph, movie/TV clips, and natural picture video sequences. The discontinuous tone content features very sharp edges and thin lines, even one-pixel thin lines while continuous tone content features relatively smooth edges and thick lines. It is well known that for continuous tone content, chroma subsampling is almost visually lossless, hence all existing video coding standards such as MPEG, H.264 & AVS are designed primarily to code chrom-subsampled contents. Actually, majority of CODEC products on the market today can handle only YUV420 or 422 format video. On the other hand, chroma subsampling significantly degrades the visual quality of discontinuous tone content in a screen image..For screen content, although the visual quality of chrom-subsampled continuous tone content oriented coders is not good, almost all applications today still uses them primarily due to two reasons: 1) the majority of video CODEC software, hardware, and human assets developed in the past 25 years are chrom-subsampled ones and completely switching to full-chroma coders (at least two coders, one for continuous tone content and the other for discontinuous tone content) not only significantly increase the unit cost but also requires tremendous new investment; 2) since quite some portions of a screen image are continuous tone content, coding the entire image using two full-chroma coders is a overkill and too expensive..Hence, the optimal solution to achieve best 3-way balance between bit-rate, coding quality and coding complexity for screen content compression is to combines one chroma sampling rate (such as full-chroma YUV444) coding with another chroma sampling rate (such as chroma-subsampled YUV420) coding and mix them in some properly selected boundaries in an adaptive way or a pre-defined way..Based on our earlier work on Mixed Chroma Sampling-rate (MCS) coding, this project studies SoC implementation, system architecture, and module-level optimal algorithm of MCS coding using the mainstream IC process for real-time ultra high-definition resolution screen content coding. An MCS coder consisting of full-chroma string-matching plus entropy coding tools united with chroma-subsampled H.264/HEVC coding tools is used as the base coder. The SoC design theory and implementation will be verified on FPGA.

云-移动计算正在成为主流信息处理和计算模式。屏幕图像编码是其中一项关键技术。由于屏幕图像含非连续色调内容如文字、图表、尖锐边界、单像素线条，对其首先色度子采样再编码的传统做法对图像质量损害极大。因此，必须对原始全色度屏幕图像直接编码。但是，全色度编解码器比色度子采样编解码器成本翻番，且需巨额新投资。因此，用混合色度采样率的编解码器对全色度屏幕图像进行压缩，既能达到高质量编码，又仅增加极小成本。前期工作中，我们提出了混合色度采样率编解码理论与算法。本项目旨在研究此类编解码器的面向超高清屏幕分辨率在主流半导体工艺下的SoC可实现性，其最优系统架构和各模块的最优算法等关键技术。本项目以全色度串匹配字典熵编码工具与色度子采样H.264/HEVC编码工具融合的混合色度采样率编解码器为基础，研究其在比特率、编码质量和复杂度三方面达到最佳综合性能的超高清屏幕用SoC设计理论与实现算法，完成FPGA验证。

基于云平台的云-移动计算正在成为主流信息处理和计算模式。“宽带中国”战略和互联网+行动计划更使中国到2020年宽带网络将基本覆盖所有城镇村乡，打通网络基础设施“最后一公里”。而屏幕内容编码，从根本上解决了基于互联网、大数据、云平台的各种应用中的数据传输瓶颈问题，使互联网所到之处，人人都能像用水和电一样以最低成本用上大数据、云计算和云平台，打通互联网+战略的“最后一米”。因此，屏幕内容编码（SCC）成为近年来受到学术界和工业界高度关注的关键技术。三大国际标准组织(ISO/ITU/IEC)在过去3年制定了最新国际标准HEVC-SCC。中国AVS工作组也已经基本完成国标兼IEEE标准AVS2SCC。本项目围绕这两大标准的制定工作，在SCC技术研究方面取得了国际最领先的成果。本项目的主要研究内容和成果之一是一整套仿2维串匹配编码算法，能提高编码效率60%，在国际上被评价为奠定了HEVC-SCC标准的基础。本项目还在串匹配算法基础上进一步提出了通用串预测算法，包括3种受限串模式，已经被AVS2SCC标准全面采纳，使AVS2SCC比HEVC-SCC的编码效率提高26%，充分显示了串匹配编码算法的优势和适用于各种不同大小和形状的匹配的优越性质。本项目的另一项主要研究内容和成果是一种通用串匹配硬件解码器的芯片系统架构理论和RTL逻辑设计与实现。这款设计以6.95万逻辑门和24KB专用SRAM的极低成本达到4K屏幕分辨率的解码，在国际上首次证明了极低成本4K超高清的通用串匹配硬件解码器的SoC芯片实现是完全可行的，解决了通用串匹配编解码技术的应用和推广中遇到的最主要问题。这些进展都具有很大科学意义。本项目的其他具体成果包括：1）8篇SCI论文（含4篇IEEE国际顶级期刊论文）和7篇EI论文，2）申请了10项中国发明专利和10项PCT国际发明专利，都有潜力进入国际、中国标准专利池，3）向HEVC-SCC标准制定工作组提交了32份技术提案，向AVS2SCC标准制定工作组提交了9份技术提案。本项目取得了丰硕的、远超出预期的、得到国际同行认可的、被国际标准和中国国标采纳的成果，也为屏幕内容编码技术和标准的进一步发展打下了坚实基础。