液晶与显示, 2015, 30 (4): 681, 网络出版: 2016-02-02   

一种多扭曲失真图像的质量评价方法

Image quality assessment algorithm for multi-distorted image
作者单位
1 长春理工大学 研究生院,吉林 长春 130022
2 长春大学 电子信息工程学院, 吉林 长春 130022
摘要
针对现有评价方法不适合于多扭曲失真图像这一问题,本文展开了对多扭曲失真图像评价方法的研究。在分析图像的边缘信息和奇异值向量对视觉特征的表征能力的基础上,提出一种基于边缘信息奇异值分解的图像质量评价算法。首先,利用Sobel算子提取参考图像及失真图像视觉敏感的边缘信息,再对两图像的边缘信息进行奇异值分解,利用奇异值向量之间的夹角来描述失真图像的畸变程度。最后,采用LIVE数据库中的450张多扭曲的失真图像验证该文算法,并与MSE、PSNR、SSIM、CSSIM等算法进行了对比。实验结果表明,该文算法对多扭曲失真图像的质量评价具有更高的稳定性,主客观评价的一致性较传统评价方法更好。通过对比时间效率,该方法基本上满足实际需求,具有更高的适用性。
Abstract
Due to the problem that the existing method of objective quality assessment algorithm is not suitable for multi-distorted images, the quality assessment method for multi-distorted images was studied in this paper.On the basis of analyzing capability that the singular vector and edge information representing visual feature, an algorithm of image quality assessment based on the singular value decomposition of edge information was put forward.Firstly, the edge information of the reference image and distorted image which is sensitive to visual information was extracted by using the Sobel operator.Secondly, the singular value of the edge information of two image was decomposed, then the angle between the singular value was calculated to describe the quality of the distorted image.Finally, the algorithm applied in this paper are verified by more than 450 distorted images in LIVE Multiply Distorted Image Quality Database, and the algorithm was compared with the MSE,PSNR,SSIM,CSSIM method.Experimental results show that the algorithm is more consistent with human subject scores and has greater stability for multi-distorted image than traditional methods.Through comparison with the time efficiency, the proposed algorithm can basically meet the practical demand, and the algorithm is more usability.

王春哲, 李杰, 李明晶, 郭盼. 一种多扭曲失真图像的质量评价方法[J]. 液晶与显示, 2015, 30(4): 681. WANG Chun-zhe, LI Jie, LI Ming-jing, GUO Pan. Image quality assessment algorithm for multi-distorted image[J]. Chinese Journal of Liquid Crystals and Displays, 2015, 30(4): 681.

本文已被 2 篇论文引用
被引统计数据来源于中国光学期刊网
引用该论文: TXT   |   EndNote

相关论文

加载中...

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!