Research On Multi-exposure Fusion Image Quality Assessment

With the rapid development of computer vision,image quality has become more and more important in many fields,such as medical imaging,remote sensing,computer vision,to name a few.Due to the limitation of the existing image acquisition equipment,it is hard to collect all the dynamic range of luminance of natural scene.Multi-exposure image fusion(MEF)is considered as an effective quality enhancement technique for this problem.The MEF aims to offering a single higher quality image with informative and clearer image details by fusing a series of images with multiple exposure levels.It not only makes up the shortage of the existing equipment,but also shows extensive application and research value in electronic and computer fields.Although many MEF algorithms have been proposed,the goal of fusion is to provide high quality images.Therefore,how to accurately evaluate the quality of MEF becomes an important research topic.Note that an effective MEF IQA metric cannot only evaluate the quality of fused images,but also serve as a guidance role in the design of image fusion algorithms.In this thesis,our focus is on the development of IQA models to evaluate the MEF images by analyzing the characteristics of MEF and HVS.We proposed a multi-scale contrast-based image quality assessment model(MCM)from the perspective of multi exposure,and other two algorithms based on the gradient fusion.The main work of this thesis can be summarized as follows:Firstly,considering the characteristics of MEF images and the multi-channel of human vision,an effective IQA model for conducting the quality of the MEF images is proposed,called MCM.The proposed MCM first extracts two important features,contrast structure and contrast saturation.For each image,the degree of similarity measured for each above-mentioned contrast attribute is then computed independently,followed by combiningthem together with weight of each reference image for obtaining contrast similarity maps.Subsequently,the image value will be achieved at single scale.Finally,a multi-scale scheme is utilized to explore the image details from finer to coarser scales for producing the final MCM score.Simulation results have shown that the proposed MCM model can well evaluate the perceptual quality of MEF.Secondly,two MEF image quality assessment algorithms are proposed based on gradient magnitude and gradient orientation,respectively.The proposed algorithm from gradient magnitude,called GMF,firstly generates a new reference image by fusion,and then calculates the gradient magnitude similarity to depict the perceptual of MEF image.Experimental results on the benchmark database have shown that the GMF produces high consistency with HVS on MEF images.Furthermore,on the basis of gradient magnitude,gradient orientation similarity is calculated and used as the weight coefficient for achieving the objective quality evaluation score of multi-exposure fusion image.The experimental results also show that the proposed GOW correlates well with subjective scores.In conclusion,the proposed algorithms in this thesis have good consistence with human perception,and low computation complexity on the assessment of the MEF image quality.This thesis is of a certain theoretical and practical value to promote the application of MEF.