Motion Deblurred Images Quality Assessment Based On Information Entropy

There are many distortions in the process of image generation,transmission and processing,and blurring is the most common one.Nowadays,many deblurring algorithms have emerged in academia,but there are no ideal assessment methods for the quality of deblurred images.In this thesis,motion deblurred images quality assessment is studied,and the high-precision quality assessment of motion deblurred images is achieved.The specific work is as follows:Firstly,the research background and significance of the thesis are introduced,and briefly review the history and researching actuality of image quality assessment methods at home and abroad and their related theories are introduced.Then,the SSEQ algorithm and its improvements are discussed.Firstly,the defects of SSEQ algorithm are analyzed,and the performance of SSEQ algorithm is improved by changing one-dimensional spatial entropy into two-dimensional spatial entropy.Secondly,a simulation model is established by using the reference images of LIVE database.Through motion blurring these reference images,eight deblurring algorithms are adopted to obtain motion deblurred images,thus adding motion deblurred images to the original LIVE database.Finally,four improved methods SSEQ_psnr,SSEQ_fsim,SSEQ_vsi and SSEQ_vif are obtained by using four full reference image assessment indexes instead of subjective scores in image database.The performance of these four methods is evaluated in MDD2013.Finally,in order to further improve the performance of the improved SSEQ algorithm,the direct evaluation methods of motion deblurred images distortions are discussed in this thesis.Combined with the improved SSEQ algorithm,we propose a new algorithm SSEMDQ for motion deblurred images quality assessment.Firstly,the three distortions of motion deblurred images and methods of estimating distortions are analyzed.By weighting their scores,the result is taken as Q_E.The evaluation value of improved SSEQ algorithm is taken as Q_S,Q_S/Q_E as the final score of the quality of the motion deblurred image,four new algorithms are obtained.Then,the performances of four new algorithms are compared in MDD2013.The results show that SROCC,KROCC,PLCC and RMSE are 0.8705,0.6830,0.8837 and 1.4004 respectively when SSEQ_vsi is used,the resulting final algorithm SSEMDQ has the best performance.Finally,the performance of SSEMDQ is analyzed from the monotony and accuracy of prediction.The experimental results show that the performance of SSEMDQ is better than the existing four kinds of representative image quality assessment methods.