基于运动剧烈程度的无参考视频质量评价模型

A No-reference Assessment Model for Video Quality Based on Movement Activity

网络视频质量评估具有无参考性、实时性、网络传输状态依赖性和主观视觉性等需求.本文以网络丢包引发的视频失真为研究重点,针对不同解码类型帧上的网络丢包引发的视频失真持续效应不同和网络丢包引发的人眼视觉感受与丢包所在帧视频内容的运动剧烈程度有着显著的关联等现象,提出了一种基于运动剧烈程度的无参考视频质量评价模型.该模型无需原始参考视频序列亦无需视频解码,对客户端接收到的码流分析其由丢包引起的视频损伤,标记受损宏块,建立受损宏块的失真持续效应和运动剧烈程度与视频质量间的关联,并完成视频质量评估.实验结果表明,该方法计算量小,实时性高,与主观评价结果一致性程度较高.

It is necessary to assess network video quality in real-time that depends on network transmission statement to subjective visual without any referenced video sequences. Focusing on distortion of videos caused by network packet loss, it is found that network packet loss in different frames with different decoding types causes different sustained effects of video distortion, and it exists significant association between visual feelings caused by network packet loss and movement activity of lost video content. Hence, a no-reference assessment model for video quality is proposed based on movement activity. This model analyzes the damage caused by the packet loss of the code stream received by client, and marks the damaged macro block to associate the damaged macro blocks＇ distortion continuing effect and movement activity with videos＂ quality. Then, it completes the quality assessment of videos without reference of any original video sequence or video decoding. Experimental results show that this method has a small amount of calculation, a high real-time and a high degree of consistency with the subjective assessment results.