Image quality assessment based on human visibility threshold theory and structural similarity
Yuanyuan Hu · JOURNAL OF SHENZHEN UNIVERSITY SCIENCE AND ENGINEERING · 2010
In the experiment on images with just noticeable distortion(JND),it was found that the influences of the luminance adaptation and contrast masking on visual perception were not taken into consideration in the image measure of structural similarity(SSIM).To overcome such limitation,the visibility threshold SSIM(VTSSIM) algorithm was proposed in the paper.First,the prototype image was divided into smooth,detailed and common image blocks according to its local standard deviations.Then,two extreme regions,i.e.,the luminance adaptation extreme regions for smooth blocks and contrast masking extreme regions for detailed blocks,were determined by the JND profile.Finally,the SSIM maps of the extreme regions together with those of the remaining blocks were weighted to get the ultimate quality assessment of VTSSIM.Experimental results show that VTSSIM is consistent with the judgments of human observers than SSIM,especially for the images with less distortion.