基于Tetrolet变换和四元数模型的视频质量评价方法  

A Video Quality Evaluation Method Based on Tetrolet Transform and Quaternion Model

作　　者：杨斌斌[1] 李纲[1] 金炜[1] 

机构地区：[1]宁波大学信息科学与工程学院,浙江宁波315211

出　　处：《移动通信》2016年第2期74-78,共5页Mobile Communications

基　　金：国家自然科学基金项目(61271399;61373068);宁波市自然科学基金资助项目(2012A610047);浙江省"信息与通信工程"重中之重学科开放基金资助项目

摘　　要：提出了一种基于Tetrolet变换和四元数的视频质量评价方法,通过对视频帧进行Tetrolet变换得到相应的纹理细节和边缘轮廓特征信息。对视频图像进行帧间残差运算后,结合结构相似性算法获取原始视频和失真视频在时域方面的信息差异,作为四元数的各个部分,再对四元数矩阵进行奇异值分解以提取特征向量的方式来获取视频评价值。对LIVE视频质量数据库进行测试,实验结果表明该方法有效地提高了视频质量评价的主客观一致性。A video quality evaluation method based on Tetrolet transform and quaternion model was proposed in this paper. The Tetrolet transform was taken for video frames to obtain the corresponding texture detail and edge contour feature information. After calculating the residual information between frames, the structural similarity algorithm was used to obtain the information difference between the original and distorted video s in time domain as all parts of quaternion. Then, the quaternion singular value decomposition was taken to extract feature vector to evaluate video. Experimental results tested on LIVE test data set showed that the proposed method effectively enhanced the consistency of subjective and objective for video quality evaluation.

关 键 词：Tetrolet变换 结构相似性 四元数奇异值分解 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]