视频质量与码率，分辨率的关系

MOS: Mean Opinion Score (MOS) 此指标用于判定视频的质量

调研总结：

1、

Thakolsri S, Khan S, Steinbach E, et al. QoE-driven cross-layer optimization for high speed downlink packet access[J]. Journal of Communications, 2009, 4(9): 669-680.

The proposed scheme aims at maximizing the user satisfaction by taking advantage of the link adaptation mechanism of HSDPA and the rate adaptation capability of multimedia applications.

本文的对于QoE的建模方面有一定的价值，它构建了一个长时间的利用率函数。本文利用MOS来衡量用户的QoE.

We use utility functions to describe the Quality of Experience (QoE) for different applications as a function of lower or radio link layer parameters, e.g., rate or throughput, time share, power, spreading code, bandwidth, etc.

U = f(R), f : R → MOS

1、对于voice call application



2、video streaming文件

the Structural SIMilarity (SSIM) index


（和视频具体的内容）

where X and Y are the reference and distorted images, respectively; xj and yj are the image contents at the jth local window; and M is the number of local windows of the image.

文章指出，信号强度和视频的质量没有必然的联系，需要用SSIM来衡量。

2、

Essaili A E, Schroeder D, Staehle D, et al. Quality-of-experience driven adaptive HTTP media delivery[C]//Communications (ICC), 2013 IEEE International Conference on. IEEE, 2013: 2480-2485.】

本文明确指出，在应用层，衡量实时的视频的质量，可以用MOS，而基本的Utility函数就可以用文1（1）的基本模型。

MOS can take on any value between 1.0 (30 dB) and 4.5 (42 dB), which represent the worst and best QoE, respectively

3、

Janowski L, Romaniak P. QoE as a function of frame rate and resolution changes[M]//Future Multimedia Networking. Springer Berlin Heidelberg, 2010: 34-45.】

本文指出，单纯地通过帧速率和图像的分辨率来判断用户体验是不够准确的，文章指出还需要结合TI（时态信息）和SI（特征信息），而这两种信息的计算方式如下：



文章提出的三个模型：



注意：t为每秒的帧数，S为图片的分辨率，非视频码率和device的分辨率