面向云计算环境的高效视频编码多粒度优化关键技术研究

项目名称： 面向云计算环境的高效视频编码多粒度优化关键技术研究

项目编号： No.61472281

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 计算机科学学科

项目作者： 王瀚漓

作者单位： 同济大学

项目金额： 82万元

中文摘要： 随着多媒体内容分享服务的迅速发展和视频显示技术的不断推陈出新，互联网上的视频数量正呈现海量级爆炸式地增长，基于高清视频和超高清视频的多媒体服务和应用正日益普及。视频大数据时代的到来为海量视频数据压缩带来了巨大挑战。高效视频编码（HEVC）作为新的视频编码标准，能更有效地提高视频数据的压缩效率，但具有庞大的编码计算复杂度。本项目旨在研究面向云计算环境的HEVC多粒度编码优化技术，从算法粒度层和资源粒度层两方面提高HEVC编码速度。在算法粒度层，研究基于多层次目标决策优化理论的HEVC快速编码算法；在资源粒度层，研究适用于HEVC的多粒度并行计算模型和调度算法，通过云节点、CPU和GPU等计算资源实现基于粗、中、细三个粒度的视频编码并行计算。本项目探索面向云计算环境的HEVC多粒度优化编码理论体系，实现视频编码的理论创新与技术突破，促进基于云计算的多媒体大数据处理理论和技术体系的研究与发展。

中文关键词： 高效视频编码；多粒度优化；云计算；GPU计算；多层次目标决策

英文摘要： With the rapid advancement of multimedia content sharing service and video display technologies, there are a vast amount of videos on the internet which are also proliferating in an exploding way, and a growing popularity is currently witnessed for the increasing demand of High Definition (HD) and even Ultra High Definition (UHD) video oriented multimedia services and applications. Therefore, the upcoming era of video big data poses great challenges to compress the vast amount of high-resolution videos. High Efficiency Video Coding (HEVC) as the newest video coding standard is designed to compress videos in a far superior rate-distortion manner, however, the requirement of huge amounts of computaitons for HEVC limits its application in real world. In this project, we are going to investigate multi-granular optimization techniques for HEVC under cloud computing environments to speed up HEVC encoding, specifically, from two optimization viewpoints including the Algorithm Granularity oriented Optimization (AGO) and the Resource Granularity oriented Optimization (RGO). Regarding AGO, we are planning to study multi-level decision making theory based methods to design fast encoding algorithms for HEVC. As far as RGO is concerned, we will explore multi-granular parallel computing models and the related scheduling algorithms, aiming to make a full use of computers, CPUs and GPUs to realize a joint parallel implementation at coarse-grained, medium-grained and fine-grained levels. This project will establish a framework of multi-granular optimization theory for HEVC and achieve the theoretical innovation and technological breakthroughs for HEVC, which will promote the development of cloud computing oriented multimedia big data processing related research fields.

英文关键词： High Efficiency Video Coding;Multi-Granular Optimization;Cloud Computing;GPU Computing;Multi-Level Decision Making