基于内容的H.264无参考视频质量评价模型  

Content-Based H.264 Non-Reference Video Quality Evaluation Model

作　　者：李晨昊[1] 卓力[1] 李嘉锋[1] LI Chen-hao;ZHUO Li;LI Jia-feng(Signal&Information Processing Laboratory,Beijing University of Technology,Beijing 100124,China)

机构地区：[1]北京工业大学信号与信息处理研究室,北京100124

出　　处：《测控技术》2019年第1期106-110,135,共6页Measurement & Control Technology

基　　金：国家自然科学基金资助项目(61531006)

摘　　要：针对H. 264视频编码标准,提出了一种基于内容的无参考视频质量评价模型。首先,从H. 264码流中提取参数,并对其进行统计分析,用于描述视频纹理复杂程度和运动剧烈程度,将其与量化参数相结合构成输入向量;然后,利用BP神经网络的方法,建立输入向量与主观评价打分(MOS)之间的映射模型,用于对H. 264视频码流的质量进行预测。实验结果表明,所提出的评价模型得到的斯皮尔曼系数(SROCC)为0.9624,皮尔逊系数(PLCC)为0.9464,可以获得较高的预测精度。A content-based non-reference video quality assessment model is proposed for the H.264 video.Firstly,the parameters are extracted from the H.264 bitstream and statistically analyzed to describe the complexity of the video texture and the intensity of the motion.Combined them with the quantization parameters to form the feature vector.Afterwards,BP network is used to establish the mapping model between eigenvectors and subjective evaluation scores(MOS)for predicting the quality of the H.264 video bitstream.The experimental results show that the Spearman coefficient(SROCC)obtained by the proposed evaluation model is 09624,and the Pearson coefficient(PLCC)is 0.9464,which can obtain higher prediction accuracy.

关 键 词：H.264 纹理丰富程度 运动剧烈程度 无参考视频质量评价 MOS 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]