计算机应用 ›› 2013, Vol. 33 ›› Issue (03): 691-694.DOI: 10.3724/SP.J.1087.2013.00691

• 多媒体处理技术 • 上一篇    下一篇

基于梯度相关性分解的无参考图像质量评价方法

廖宇*, 郭黎   

  1. 湖北民族学院 信息工程学院,湖北 恩施 445000
  • 收稿日期:2012-09-17 修回日期:2012-11-14 出版日期:2013-03-01 发布日期:2013-03-01
  • 通讯作者: 廖宇
  • 作者简介:廖宇(1979-),男,湖北建始人,讲师,硕士,主要研究方向:信号与信息处理; 郭黎(1978-),女,湖北浠水人,讲师,博士研究生,主要研究方向:信号与信息处理。
  • 基金资助:

    国家自然科学基金资助项目(61263030); 湖北省自然科学基金资助项目(2012FFC02601); 湖北省教育厅科学研究项目(Q20111907)。

New no-reference image quality assessment method based on decomposition of gradient similarity

LIAO Yu*, GUO Li   

  1. School of Information Engineering, Hubei University for Nationalities, Enshi Hubei 445000, China
  • Received:2012-09-17 Revised:2012-11-14 Online:2013-03-01 Published:2013-03-01
  • Contact: LIAO Yu

摘要: 目前大部分无参考型的图像质量评价方法都是基于图像的几何特征进行描述的,但是这种方法对于图像的边界要求较为严格,并且在实际应用中的图像的失真类型是未知的。针对这一缺点,提出一种基于梯度相关性分解的无参考图像质量评价(DGS)方法,该方法提取图像的梯度,对其进行奇异值分解作为图像的主要结构信息,以此对图像的质量进行评价。实验结果表明,DGS模型比通用的简单有效的峰值信噪比(或均方误差)模型更符合人眼视觉系统特性,能在无参考的情况下更好地评价图像质量,并与图像的主观评价值达到更准确的一致性。

关键词: 无参考图像质量评价, 梯度结构相似度, 人眼视觉系统, 奇异值分解

Abstract: At present, the existing general no-reference image quality assessment methods are used for special purpose. Furthermore, the distortion type of the image is unknown in practical application. Most of no-reference image quality assessment methods are based on the geometrical description feature of image, but this type of method relies on the boundary of image too much. To solve this problem, a no-reference image quality assessment method was proposed based on the gradient similarity decomposition of image, which was named Decomposition of Gradient Similarity (DGS). The proposed method extracted the gradient feature and decomposed the correlation of gradient as the main structure information of the image. The experimental results show that the proposed DGS model is better than Peak Signal-to-Noise Ratio (PSNR) (or Mean Square Error (MSE)) model, which is more sensitive to Human Visual System (HVS) characteristics and more consistent to the subjective evaluation value.

Key words: no-reference image quality assessment, gradient structure similarity, Human Visual System (HVS), Singular Value Decomposition (SVD)

中图分类号: