基于人眼视觉特性与SVM的视频质量评估模型

Human eye system and SVM-based video quality assessment model

建立一个精确、有效的视频质量评估模型是当前多媒体研究的热点问题。通过分析现有无参考视频质量评价技术的优缺点,文中提出了基于人眼视觉特性(human visual system,HVS)与SVM的视频质量评估模型——HVQAM模型。该模型在采用支持向量机对帧图像进行质量评估的基础上,创新地采用人眼视觉特性对帧内区域图像进行加权平均。通过实验对HVQAM模型进行了测试,测试结果表明文中所提出的模型与主观评估结果相比较,相似度更高,而CPU及内存占用率较低。

It has become a hot issue in current researches that how to establish an effective model to evaluate video quality accurately in real time. By analyzing the advantages and disadvantages of existing without reference video quality evaluation technologies,this paper proposes a human eye system and SVMbased video quality assessment model-HVQAM. The model can evaluate the frame image quality based on support vector machine and apply the human visual characteristics of innovation to weighted average of frame image area. The HVQAM is tested by an experiment. Experimental results show that evaluation results of the presented model and subjective evaluation results are more similar,and the CPU and memory usage rate of the model is lower.