Abstract: In order to solve the problem of video distortion caused by video compression, this paper proposes a no-reference Video Quality Assessment（VQA） model that utilizes many spatial and frequency features based on the analysis of perceptual feature of video’s quality. This method mainly extracts spatial and frequency perceptual features, including Gray-level Gradient Co-occurrence Matrix（GGCM）, spatial entropy, spectral entropy, correntropy and a natural index features. In the process of extracting the video features, this method calculates the variance of video frame as the whole video features, which has a better performance on distinguish between different types of distortion of the video than the traditional ways. Finally, these features are trained by Support Vector Regression（SVR） model to build the relationship with perceptual features and quality of distorted video. The proposed model is tested on LIVE VQA database and on IVP VQA database and experiment results prove that the proposed algorithm can achieve much higher consistency with the subjective evaluation than state-of-the-art published algorithm.

Key words: spatial entropy, spectral entropy, support vector regression, video quality assessment

摘要： 针对视频压缩等处理导致视频失真的问题，通过对视频质量感知特征的分析，提出一种空域和频域联合特征挖掘的无参考视频质量评价方法。该方法主要提取了空域和频域联合感知特征，包括灰度-梯度共生矩阵、空间熵、谱熵、相关熵以及自然指数特征。在提取视频特征的过程中，通过计算视频帧特征方差来表示整个视频的特征，比传统方法中取视频帧平均值更有利于区分不同失真类型的视频。最后，使用支持向量回归模型构建了感知特征与视频质量之间的关系。该方法在LIVE和IVP 视频数据库上的实验结果表明，提出的方法相较当前文献报道方法，有着更好的性能。

关键词: 空间熵, 频谱熵, 支持向量回归, 视频质量评价