基于视觉质量评价的高效三维视频编码研究

项目名称： 基于视觉质量评价的高效三维视频编码研究

项目编号： No.61472388

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： 齐洪钢

作者单位： 中国科学院大学

项目金额： 82万元

中文摘要： 三维（3D）视频编码技术发展面临很多挑战，如：编码效率不高，视觉质量评价不精确，深度信息重建质量不理想等。必须有效的解决这些问题，3D视频才能获得实际的应用。本课题在对视频编码理论深入研究基础上，针对3D视频编码技术中存在的问题，1)研究率失真优化理论，分析视频和深度图编码的不同率失真行为，建立视频和深度图加权的率失真编码框架，提高3D视频编码性能。2)研究立体视觉质量评价方法，分析深度和平面视频的视觉响应，探索平面视频与深度信息对视觉感知的共同作用，建立平面视频和深度联合视觉质量评价模型，准确评估3D视频的视觉质量。3)研究深度信息视觉重建技术,分析深度图编码信息损失机理，结合视频的深度感知，提供深度重建精确信息，全面考虑主观质量和客观质量评价效果，建立基于主客观质量联合自适应评价的深度重建优化模型，重建高视觉质量的深度图。本课题的研究成果可为3D视频编码应用提供理论和技术支持。

中文关键词： 三维视频编码；率失真优化；视觉质量评价；深度信息重建

英文摘要： The development of three Dimension (3D) Video coding Technologies is facing many challengs. For example, coding effeciency is not high, visual quality assessment is not accurate, reconstructed quality of depth is not satisfied and so on. It is impossible that 3D video can be applied in practice, until these problems have to be efficiently solved. Based on thorough research on the theories of video coding, for sloving the existing probles on 3D video coding technology, some concrete researches will be done in this project. 1) Rate Distortion (R-D) optimization theroy is studied in this project. The different R-D behaviors of coding video and depth are analyzed. The R-D optimization framework for both video and depth is setup for achieving the global optimal coding performance. 2) The visual quality assessment method of 3D video is also studied in this project. Visual perception to depth and 2D video is analyzed. The joint stimulus of 2D video and depth signals to perception is explored. The visual quality assessment model of joint video and depth is established for the accurately visual assessment of 3D video. 3) Moreover, visual depth reconstruction is studied in this project. The mechanism of lossy video coding is analyzed for obtaining the accurate depth reconstruction information through the video aware. The effects of both subjective and objective visual qualities are considered for setting up the subjective and objective qualities jointly adaptive assessment based optimization model for visual depth reconstruction. With the model, depth map are reconstructed in high visual qualtiy. The achievements of this project can provide supports of theory and technology for applications of 3D video coding.

英文关键词： 3D Video Coding;Rate Distortion Optimization;Visual Quality Assessment;Depth Reconstruction