Research On Image Quality Assessment Method Based On Visual Perception

In today's life,digital image as a tool of information expression and communication has been widely used.However,in the process of image acquisition,analysis,transmission,processing and reconstruction,it is easy to cause image distortion for a variety of reasons,affecting the visual effect of the image itself.How to keep the image quality consistent with the human visual effect is the core con tent of image quality assessment research.Based on the previous research on image quality assessment model,this thesis will explore the perceptual characteristic s more in line with the human visual system,as well as the statistical characteristics of natural scenes,in order to extract features that can reflect certain characteristics of the image,so as to achieve a higher consistency between objective assessment and subjective assessment.(1)Aiming at the full reference model,an image quality assessment method based on HSV color space and edge features is proposed.Firstly,the image is transformed into HSV color space which accords with human visual characteri stics,and the brightness components and color features are obtained.Then,according to the characteristics of human visual masking,the brightness component and the edge of gray image are extracted as local features of slow and strong changes,respectively.Then visual saliency model is used to obtain the visual quality of strong change area,slow change area and color feature as complementary edge extension area.Finally,the visual saliency model is used as a weighting function to fuse the visual quality of the local area,so as to determine the final image quality score.The experimental results on four common open databases(LIVE,CSIQ,TID2008,TID2013)show that the assessment index of the proposed algorithm has achieved relatively ideal results on the whole,and SROCC index is significantly higher than other algorithms in different distortion types of TID2013 database.(2)Aiming at the no-reference model,an image quality assessment method based on structure and natural scene statistics is proposed.Firstly,according to the characteristics of human visual system,the edge and texture of the image are extracted as the perceptual features of the image.Then the image is transformed in to the Lalpha beta color space which is more suitable for natural image processing.The brightness components of the image and the red-green and blue-green components as color features are extracted as statistical features of natural scenes.Then the brigh tness and color space statistical parameters are obtained by using the asymmetric generalized Gauss model.Finally,the extracted features are input into the support vector regression model to predict the image quality.The experimental results on common d atabases(LIVE,CSIQ,TID2008,TID2013)show that the proposed algorithm can obtain more consistent results with subjective assessment.