Research On Image Quality Evaluation Method Based On Visual Saliency Weighting

At the beginning of this century,as a tool for information exchange and expression,digital image processing technology has developed rapidly and has been widely used in most areas such as intelligent transportation,smart cities,urban security,biomedicine,and military research.After many experts and scholars make unremitting efforts and persist in exploring,improve and perfect image quality evaluation methods,and establish evaluation models.However,it is still difficult to guarantee the quality of digital images that have only undergone preliminary processing or have not yet been processed.In Image Acquisition,Image Transmission,Image Analysis,Image Processing,and Image Reconstruction,distortion is very likely to occur.Second,with the increase in monitoring equipment,the amount of image data has also grown exponentially.In order to better ensure the collection of high-quality images,improve the utilization of image information,improve the work efficiency of the monitoring system,reduce the number of invalid image information,and accurately quantify and evaluate the quality of images efficiently,it has become one of the research hotspots.For this reason,it has become an urgent need to further improve the existing image quality model and conduct in-depth research on the high image quality evaluation system in order to develop a more efficient system.This article focuses on research and improvement of image quality assessment algorithms.Its content is as follows:1.Define the image and image quality.The image quality assessment methods were elaborated,including subjective image quality assessment and objective image quality assessment.Focusing on the analysis of various methods of objective image quality assessment,a comprehensive analysis of the full reference image quality evaluation system was made.Point out where the various algorithms need to be optimized and simplified.2.Conduct an in-depth study of the human visual system to analyze the physiological structure and working principle of the human visual system.Summarizes the four low-level psychophysical properties of the human visual system,including luminance characteristics,contrast sensitivity characteristics,mask characteristics,and foveal visual characteristics.It provides a theoretical basis and foundation for the modeling and experimentation of subsequent image quality assessment algorithms.3.For most of the current saliency algorithms with high complexity,and for the extraction of significant regions mixed with a large number of human factors and other issues,this paper proposes a saliency algorithm based on local convex rectangular region of interest,the algorithm to human The fovea visual characteristics of the visual system are the theoretical basis,and the significance region of the image is extracted using a random extraction method.Experiments have shown that the algorithm has a good consistency with the human visual system and has achieved ideal results.4.Aiming at the phenomenon that the mature SSIM algorithm is less effective in evaluating fuzzy images,a visually significant weighted image quality evaluation model is proposed and verified in the LIVE image library.After experimental demonstration,this method has significantly improved than the traditional PSNR,SSIM and GSSIM algorithms.Finally,the application of the image captured by the highway was used to verify the applicability of the algorithm.Experimental results show that this method can effectively evaluate various types of distorted image quality.The evaluation results have certain consistency with the human visual system and have high accuracy.