深度视频编码快速模式选择算法研究

发布时间：2018-11-04 17:35

【摘要】：3D视频能够为观众带来更真实的立体感知,在电影电视等领域受到了广泛的关注。MVD(Multi-view Video Plus Depth)是当前主要的3D视频格式之一。它在彩色视频格式的基础上加入了深度视频,从而可以通过基于深度图像的绘制技术绘制已编码视点间的虚拟视点。MVD视频在压缩彩色视频的基础上,需要对深度视频进行压缩。传统的H.264/AVC标准和最新的高效视频编码标准(HEVC)都提出了较有效地压缩编码方法,但同时又带来了巨大的编码复杂度。为了解决这一问题,许多学者提出了各种快速模式选择算法。考虑到深度视频有着与彩色视频不同的特性,本文针对多视点深度视频编码过程中的模式选择问题进行了研究,并着重探索了视点间相关性和分量间相关性对深度视频模式选择快速算法的影响。传统的H.264/AVC将视频的每一帧平均划分为若干16×16的宏块,以宏块为单位逐个进行运动估计等编码过程。对于运动估计,H.264/AVC也给出了多种帧间、帧内预测模式,并通过遍历所有模式以及比较率失真代价,得到最佳模式。模式选择所带来的编码复杂度是本文关注的重点。考虑到MVD格式中深度视频具有视点间相关性,深度视频彩色视频间具有分量间相关性,本文结合两种相关性提供的先验信息,基于JMVC平台,提出了一种基于视点间与分量间相关性的深度视频快速模式选择算法。实验结果表明,本文所提出的算法可以在保证视频质量的前提下有效地提高编码速度。3D-HEVC是一种支持MVD视频格式的新兴标准,也是HEVC的一种扩展。它采用了HEVC中的编码结构,将编码块划分为编码树单元,对每一个编码单元都要进行预测编码。预测编码过程中仍存在模式选择过于复杂的问题。本文对于HEVC中的MERGE模式进行研究,并在HTM平台上探索了参考视点间、分量间相关性的MERGE模式提前截止方法。实验结果表明,所提出的方法可以在视频质量基本不变的条件下减少编码时间。
[Abstract]:3D video is one of the main 3D video formats, and has attracted wide attention in the field of film and television. It adds the depth video to the color video format, so it can draw the virtual view between the coded views by the depth image-based rendering technology. MVD video is based on the compressed color video. The depth video needs to be compressed. Both the traditional H.264/AVC standard and the latest efficient video coding standard (HEVC) have proposed a more effective compression coding method, but at the same time, it has brought huge coding complexity. In order to solve this problem, many scholars have proposed a variety of fast pattern selection algorithms. Considering that depth video has different characteristics from color video, this paper studies the mode selection in the process of multi-view depth video coding. The effects of interview correlation and inter-component correlation on the fast algorithm of depth video mode selection are discussed. In traditional H.264/AVC, each frame of video is divided into several 16 脳 16 macroblocks on average, and motion estimation and other coding processes are carried out one by one in macroblock units. For motion estimation, H.264/AVC also gives several inter-frame and intra-frame prediction modes, and obtains the best mode by traversing all modes and comparing rate-distortion costs. The coding complexity brought by pattern selection is the focus of this paper. Considering that depth video in MVD format has inter-view correlation, and color video in depth video has inter-component correlation, this paper combines the prior information provided by two kinds of correlation, based on JMVC platform, A fast mode selection algorithm for depth video based on the correlation between viewpoints and components is proposed. Experimental results show that the proposed algorithm can effectively improve the coding speed while guaranteeing video quality. 3D-HEVC is a new standard supporting MVD video format and an extension of HEVC. It adopts the coding structure of HEVC, divides the coding block into coding tree units, and predication and encode each coding unit. In the process of prediction coding, the problem of mode selection is still too complicated. In this paper, the MERGE pattern in HEVC is studied, and the MERGE pattern cut-off method based on HTM platform is explored for the correlation between reference viewpoints and components. The experimental results show that the proposed method can reduce the coding time under the condition that the video quality is almost invariable.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN919.81

【相似文献】