Research On Algorithms Of Blind Image Quality Assessment

Most information acquired by humans comes from visual systems,but the quality of digital images may be destroyed in various processes.Image quality assessment has become a necessary and important cornerstone during the development of digital image technology.In order to evaluate image quality automatically under the condition that most of original images may be unavailable,blind image quality assessment has become one of research hotspots.In this paper,gradient operators,texture features,histogram features and so on are applied based on the existing methods on image quality assessment.Meanwhile,the methods of regression analysis,Bayes classifier,decision tree classifier and others are applied on the research of image quality assessment.After a lot of experiments and theoretical studies,the blind sharpness assessment methods GI-F and SINI are successively proposed.The latter is about 3.53%higher than the latest algorithm ARISM_C on consistency performance.The accuracy of the proposed JPEG distortion detection method is 0.9447.In the implementation of algorithms,this paper focuses on cache optimization and instruction sets technology to improve runtime performance.The system Android as an example is taken to introduce the implementation of image quality assessment methods on terminals.About desktop,the design of system and software,the algorithm modules,algorithm performance,and the experimental results are detailed respectively in this paper.Finally,related work of this paper is summarized and the further research directions about blind image quality assessment in the future are pointed out.