3D视频编码的比特分配及码率控制研究
发布时间:2018-08-22 08:18
【摘要】:与传统的2D视频相比,3D视频通过对立体视觉场景的再现,使观察者体验到更加自然逼真的场景信息。作为3D视频处理的关键技术之一,3D视频的压缩对其应用具有重要意义。比特分配和码率控制技术是3D视频编码的重要组成部分。根据不同的视频编码标准和实际应用,研究者们提出了许多有效的码率控制算法,但是这些算法主要针对单个彩色视频编码的码率控制。与单个彩色视频编码相比,3D视频的编码更为复杂,从而导致现有的码率控制算法不适用于3D视频编码,无法实现视点间、彩色与深度视频间的比特分配。因此面向3D视频编码的比特分配和码率控制算法的研究具有重要的意义。为此,本文对3D视频编码中的比特分配和码率控制进行了深入的研究,主要包括以下两个方面:1)通过对虚拟视点失真的分析,本文提出了一种基于指数D-R模型的彩色与深度视频联合编码的比特分配和码率控制方案。通过最小化虚拟视点的失真,该方案将最佳比特分配的问题转化为一个约束条件下求极值的数学问题并通过拉格朗日乘数法求解。所提出的比特分配及码率控制方案允许彩色视频和深度视频单独采用码率控制进行编码,既简化计算,又提高了最优解的精度。实验结果表明,该方案可以较准确地控制比特分配,满足3D视频系统的应用要求。2)考虑到深度视频在3D视频中的重要的地位,从提升虚拟视点质量的角度着手,本文提出了一种基于区域的深度视频编码的比特分配和码率控制算法。首先,提出一种基于虚拟视点失真分析的深度视频区域划分方法,将深度视频划分为纹理区和平滑区;然后,给出了基于区域的指数D-R模型和指数QP-R模型,用以模拟区域RD性能;最后,基于线性虚拟视点失真模型,实现基于区域的最佳比特分配和码率控制方案。实验结果表明,所提出的方法能提高深度视频编码目标码率的准确性和绘制的虚拟视点的质量。
[Abstract]:Compared with the traditional 2D video, 3D video can make the observer experience more natural and realistic scene information by reproducing the stereo vision scene. As one of the key technologies of 3D video processing, 3D video compression plays an important role in its application. Bit allocation and rate control are important components of 3D video coding. According to different video coding standards and practical applications, many effective rate control algorithms are proposed, but these algorithms are mainly for single color video coding rate control. Compared with the single color video coding, the coding of 3D video is more complex, so the existing rate control algorithm is not suitable for 3D video coding, and can not realize the bit allocation between view point, color and depth video. Therefore, the research of bit allocation and rate control algorithms for 3D video coding is of great significance. Therefore, in this paper, the bit allocation and bit rate control in 3D video coding are deeply studied, including the following two aspects: 1) through the analysis of virtual view distortion, This paper presents a bit-allocation and bit-rate control scheme for color and depth video coding based on exponential D-R model. By minimizing the distortion of virtual view, the problem of optimal bit allocation is transformed into a mathematical problem of finding extreme value under constraint conditions and solved by Lagrange multiplier method. The proposed bit-allocation and bit-rate control scheme allows color video and depth video to be coded separately using bit-rate control, which not only simplifies the calculation but also improves the accuracy of the optimal solution. Experimental results show that the proposed scheme can control bit allocation more accurately and meet the application requirements of 3D video system. 2) considering the important position of depth video in 3D video, we can improve the quality of virtual view. In this paper, a bit allocation and rate control algorithm for deep video coding based on region is proposed. Firstly, a method of depth video region partition based on virtual view distortion analysis is proposed, which divides depth video into texture region and smooth region, then, the exponential D-R model and exponential QP-R model based on region are given. Finally, based on the linear virtual view distortion model, the optimal bit allocation and bit rate control scheme based on the region are realized. The experimental results show that the proposed method can improve the accuracy of the target rate of depth video coding and the quality of the virtual view.
【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
本文编号:2196528
[Abstract]:Compared with the traditional 2D video, 3D video can make the observer experience more natural and realistic scene information by reproducing the stereo vision scene. As one of the key technologies of 3D video processing, 3D video compression plays an important role in its application. Bit allocation and rate control are important components of 3D video coding. According to different video coding standards and practical applications, many effective rate control algorithms are proposed, but these algorithms are mainly for single color video coding rate control. Compared with the single color video coding, the coding of 3D video is more complex, so the existing rate control algorithm is not suitable for 3D video coding, and can not realize the bit allocation between view point, color and depth video. Therefore, the research of bit allocation and rate control algorithms for 3D video coding is of great significance. Therefore, in this paper, the bit allocation and bit rate control in 3D video coding are deeply studied, including the following two aspects: 1) through the analysis of virtual view distortion, This paper presents a bit-allocation and bit-rate control scheme for color and depth video coding based on exponential D-R model. By minimizing the distortion of virtual view, the problem of optimal bit allocation is transformed into a mathematical problem of finding extreme value under constraint conditions and solved by Lagrange multiplier method. The proposed bit-allocation and bit-rate control scheme allows color video and depth video to be coded separately using bit-rate control, which not only simplifies the calculation but also improves the accuracy of the optimal solution. Experimental results show that the proposed scheme can control bit allocation more accurately and meet the application requirements of 3D video system. 2) considering the important position of depth video in 3D video, we can improve the quality of virtual view. In this paper, a bit allocation and rate control algorithm for deep video coding based on region is proposed. Firstly, a method of depth video region partition based on virtual view distortion analysis is proposed, which divides depth video into texture region and smooth region, then, the exponential D-R model and exponential QP-R model based on region are given. Finally, based on the linear virtual view distortion model, the optimal bit allocation and bit rate control scheme based on the region are realized. The experimental results show that the proposed method can improve the accuracy of the target rate of depth video coding and the quality of the virtual view.
【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
【参考文献】
相关期刊论文 前1条
1 朱仲杰;梁丰;蒋刚毅;郁梅;;面向立体视频的比特分配与码率控制算法[J];通信学报;2007年07期
,本文编号:2196528
本文链接:https://www.wllwen.com/kejilunwen/wltx/2196528.html