Research On Two Methods Of Full-Reference Color Image Quality Assessment

Almost all images are presented in color,but most of the current Image Quality Assessment(IQA)algorithms are mainly designed for grayscale images.The color information of the image will affect people's perception of image quality,especially it is in the case where the image is subject to color distortion.Most current literatures on color image quality assessment(CIQA)views the R,G,B channels of a color image as three separate grayscale images,or be converted to a visually consistent color space.And their IQA scores are averaged as a score for evaluating color image quality,but this method ignores the complexity of the human visual system(Human Visual System,HVS)perception of color.Although the luminance component of a color image contains a large amount of visual information,it always results in the loss of some color information during the color image processing.Therefore,it is very important to seek a CIQA algorithm that is consistent with HVS and takes into account the loss of color information.In view of this,we mainly study the two methods of full reference CIQA in this paper.The main work of the paper is as follows:(1)A full-reference CIQA method for dimension reduction is proposed.Because the color image produces high computational complexity when evaluating image quality,a CIQA model with dimensionality reduction is designed using the decolorization method.Firstly,the dimensionality of the color image is reduced by the parameterization technique,while the contrast information in the grayscale image is preserved.Secondly,a color-to-gray structure similarity algorithm(CTG-SSIM)is constructed.IQA method,Color-to-gray Gradient Weighted Lifting Structure Similarity Algorithm CTG-GWL-SSIM IQA method,based on Color-to-gray Visual Saliency-Induced Index,CTG-VSI)IQA method.A large number of experiments performed on two standard image quality evaluation databases show that the three proposed CIQA methods are competitive with state-of-the-art methods.(2)An improved CIQA method for fuzzy structural similarity is proposed.Commonly used pixel similarity measures such as mean absolute error,peak signal-to-noise ratio,mean square error and normalized color difference,cannot be well matched with visual perceptual similarity Considering that the generalized gradient can describe the structure information of the image effectively,HVS has nonlinear characteristics for image quality perception and the evaluation of different components of the image is different,this paper combines the general gradient and similarity of fuzzy structure to propose a fuzzy structure similarity CIQA method.Experimental results show that the proposed method can be competitive with state-of-the-art methods in terms of consistency with subjective evaluation.There are obvious advantages for color images with low correlation between different channels.