Journal of South China University of Technology (Natural Science Edition) ›› 2017, Vol. 45 ›› Issue (3): 11-19.doi: 10.3969/j.issn.1000-565X.2017.03.002

• Electronics, Communication & Automation Technology • Previous Articles     Next Articles

Feature Similarity Image Quality Assessment on the Basis of Human Visual System

SUN Yan-jing LIU Dong-lin XIE Xin-xin WANG Yan-fen   

  1. School of Information and Control Engineering,China University of Mining and Technology,Xuzhou 221116,Jiangsu,China
  • Received:2016-06-17 Revised:2016-09-07 Online:2017-03-25 Published:2017-02-02
  • Contact: 孙彦景( 1977-) ，男，博士，教授，主要从事图像处理、无线传感器网络和信息物理系统研究． E-mail:327724248@qq.com
  • About author:孙彦景( 1977-) ，男，博士，教授，主要从事图像处理、无线传感器网络和信息物理系统研究．
  • Supported by:

    Supported by the National Natural Science Foundation of China( 51274202) ， the National Natural Science Foundation of China for Young Scientists ( 51504255，51504214) ， the Transformation Program of Scientific and Technological Achievements of Jiangsu Province of China( BA2012068) ， the Natural Science Foundation of Jiangsu Province of China( BK20131124) ，the Natural Science Foundation of Jiangsu Province of China for Young Scientists( BK20130199) ，the Perspective Research Foundation of Production Study and Research Alliance of Jiangsu Province of China( BY2014028-01) and the Fundamental Research and Development Foundation of Jiangsu Province( BE2015040)

PDF

303

Abstract:

As the existing image quality evaluation methods of feature similarity ( FSIM) is inefficient in image information uncertainty measurement and edge information detection,a novel algorithm named HFSIM is proposed on the basis of the internal generative mechanism of human visual system ( HVS) .In this algorithm,the auto-regressive ( AR) model is employed to decompose distorted images,and the original image is decomposed into two portions,one is the predicted portion and the other is the disorderly portion.By combining FSIM with edge structural similarity ( ESSIM) algorithm,the predicted portion of image is measured,and,by employing the multi-scale peak signal-to-noise ratio ( PNSR) ,the distortion of the disorderly portion is measured.Finally,the overall image quality score is obtained according to the above-mentioned measured results of the predicted and the disorderly portions.It is found from the experiments on six public benchmark databases that the proposed algorithm is highly consistent with human perception,and that it possesses high performance in the assessment of different types of distorted images.

Key words: image quality assessment, human visual system, internal generative mechanism, feature similarity, edge structural similarity

CLC Number: