基于时域梯度相似度的视频质量评价模型  

Video Quality Assessment Model Based on Gradient Similarity of Temporal Domain

作　　者：张淑芳[1] 韩泽欣 张聪[1] ZHANG Shu-fang;HAN Ze-xin;ZHANG Cong(School of Electronic Information Engineering,Tianjin University,Tianjin 300072,China)

机构地区：[1]天津大学电子信息工程学院

出　　处：《控制工程》2018年第8期1355-1359,共5页Control Engineering of China

基　　金：国家自然科学基金青年基金资助项目(61101226)

摘　　要：视频质量评价算法有助于多媒体网络系统优化和视频编解码算法改进,近年来已成为图像质量评价领域的热门研究方向。在图像质量评价模型（FSIM）的基础上,结合视频局部多帧之间的时域相关性,通过采用新型的三维梯度算子计算原始视频序列与失真视频序列间的时域梯度相似度,提出一种基于时域梯度相似度的视频质量评价模型（VQAM-GSIM）。在LIVE视频数据库上的测试结果表明,提出的模型与视频主观评价有较好的一致性,SROCC与PLCC指标优于VSSIM和VQM两种广泛使用视频质量评价算法。Video quality assessment algorithms, which can be useful in improving the multimedia network system and video codec algorithm, have become a popular research topic in the image quality assessment field. In this paper, a novel video quality assessment model is proposed based on gradient similarity of temporal domain（VQAM-GSIM）. And this model calculates the gradient similarity index between reference videos and distorted videos by a new 3-D gradient operator on the basis of a feature similarity index for image quality assessment（FSIM） and the dependencies between adjacent frames. Experimental results on LIVE database have shown that the proposed method is better than other representative methods, while having low computational complexity and good robust performance.

关 键 词：视频质量评价 时域相关性 三维梯度算子 梯度相似度 

分 类 号：TP37[自动化与计算机技术—计算机系统结构]