视频编码中视觉质量一致性评价及其控制算法的研究

In video quality assessment, the consistent visual quality of video is very important to the visual perception of viewers. However, the traditional rate control algorithms of video coding are not efficient in visual quality consistence because there is not a mechanism of visual quality control in traditional rate control algorithms. This project firstly studies the visual quality consistence assessment of videos on the basis of the subjective and objective visual quality assessment of image and video. An objective metric is proposed to assess the visual quality consistence. Secondly, to control the visual quality consistence of videos, a theoretical model which constructs the relationship among channel bandwidth, buffer capacity and visual quality consistence is proposed based on the Law of Large Numbers and the Central Limit Theorem. This model could dynamically assess and control the visual quality consistence during encoding process. Thirdly, the algorithms based on the theoretical model are investigated for benefiting the visual quality consistence of video encoding. By introducing the proposed metric of visual quality consistence into rate control, a novel rate control algorithm is proposed under the constraints of channel bandwidth, buffer capacity and visual quality variation bound. The output of this project improves the existing rate control algorithms significantly in visual quality control. In addition, it also contributes a good technique to video streaming applications.

在视频质量评价中，视频图像质量一致性，即视频图像质量波动是影响用户视觉感受的一个非常重要的因素。而传统码率控制算法缺少质量控制的机制，所以很难保证编码图像质量的一致性。本项目首先从视频图像质量的主、客观评价入手，提出视频图像质量一致性的客观评价指标。其次，为了实现视频图像质量一致性的控制，从误差分析的角度出发，结合大数定律、中心极限定理提出了视频图像质量一致性控制理论模型，建立了带宽、缓冲区、图像质量波动限制和模型可控参数之间的关系，从而实现编码过程中视频图像质量的动态调控。再次，基于上述理论模型，探索视频图像质量一致性的控制算法，引入上述视频图像质量一致性评价指标，研究在新的质量波动指标限制和已有带宽、缓冲区限制条件下的新型的码率控制算法。研究成果显著地改善现有码率控制算法在质量控制方面的性能，并对视频流媒体应用提供了更好的技术支持。

在视频质量评价中，视频图像质量一致性，即视频图像质量波动是影响用户视觉感受的一个非常重要的因素。而传统码率控制算法缺少质量控制的机制，所以很难保证编码图像质量的一致性。本项目首先从视频图像质量的主、客观评价入手，提出视频图像质量一致性的客观评价指标。其次，为了实现视频图像质量一致性的控制，从误差分析的角度出发，结合大数定律、中心极限定理提出了视频图像质量一致性控制理论模型，建立了带宽、缓冲区、图像质量波动限制和模型可控参数之间的关系，从而实现编码过程中视频图像质量的动态调控。再次，基于上述理论模型，探索视频图像质量一致性的控制算法，引入上述视频图像质量一致性评价指标，研究在新的质量波动指标限制和已有带宽、缓冲区限制条件下的新型的码率控制算法。研究成果显著地改善现有码率控制算法在质量控制方面的性能，并对视频流媒体应用提供了更好的技术支持。