Image Quality Assessment Satisfying Subjective And Objective Consistency For Degraded,enhanced And Restored Images

Image quality assessment technology is widely used in the fields of image filtering,video quality monitoring,processing algorithm evaluation,transmission system parameter optimization and so on.This paper focuses on the research of full reference image quality assessment method satisfying subjective and objective consistency cater to degraded enhanced and and restored imagesA degraded image quality assessment method based on saliency image edge is proposed.According to the human visual attention to the content of the high frequency information such as image edge,texture and structure,the saliency of image is divided into color saliency,frequency saliency and position saliency.Finally calculating the Hamming distance of the edge detection map between reference image and distortion image edge to get the image quality assessment index EVSH(Edge of Visual Saliency image Hamming distance).The experimental results show that the proposed index has a high degree of subjective and objective consistency,and has a good evaluation effect on the distortion images.An enhanced image quality assessment method based on multiple features combination is proposed.According to the influence of image enhancement method on enhanced image quality and visual effect,the Gauss window function is used to calculate the four similarity features(brightness,contrast,sharpness,saturation)between reference image and enhanced image.These four similarity features are combined and the appropriate weights are selected by the control variable method to get the enhanced image quality assessment index EIQA(Enhanced Image Quality Assessment).The experimental results show that the proposed index has a high degree of subjective and objective consistency,and has a good evaluation effect on the enhanced images.A restored image quality assessment method based on the difference of wavelet domain information entropy is proposed.Firstly,the reference image and the restored image are decomposed by using steerable pyramid,and the images are transformed from spatial domain to wavelet domain.Secondly,the Gauss scale mixture model and the visual distortion model are used to model the reference image and the restored image in the wavelet domain.Then the coefficients of each subband are obtained.The best sub-band is selected by experiment comparison,and finally the wavelet domain information entropy difference between the reference image and the restored image is calculated to get the restored image quality assessment index RIQA(Restored Image Quality Assessment).A restored image database(RIDB)was created.The experimental results show that the proposed index has a high degree of subjective and objective consistency,and has a good evaluation effect on the restored images.