Research On Quality Index Of Image Fusion

Image fusion technology can synthesize complementary information between different sensor images,restrain redundant information between images and improve information utilization.In order to evaluate the rationality and superiority of various fusion algorithms,it is necessary to evaluate the quality of fusion images.At present,the quality evaluation of fused image is mainly from subjective and objective perspectives.Subjective quality evaluation has the characteristics of direct and accurate,but the evaluation efficiency is low and difficult to reproduce.The objective quality evaluation index can effectively evaluate the quality of fused image.Due to the lack of ground truth in image fusion,the objective quality evaluation of fused images has become a major problem in current research,and it is also one of the urgent problems to be solved in the field of image fusion.In this thesis,several typical application scenarios of image fusion are studied,and a new fusion image quality index is proposed.The main contents of this thesis are as follows:(1)Most of the fusion scenarios are observed by human eyes.In this thesis,a framework for subjective evaluation experiment of image fusion is built to verify the rationality of objective quality indicators.In the experiment,the current classical fusion method is selected to fuse images of different scenes,and the fusion results are taken as observation samples.The experiment invites teachers and students engaged in relevant research to study the subjective evaluation criteria,and then according to the criteria,combined with their subjective feelings,the selected fusion results are subjectively evaluated and scored.The experimental results will be used to validate the rationality of the following objective quality indicators.(2)The widely used image fusion quality indicators are studied,and the existing problems in the indicators are discussed.The correction methods are proposed to make the quality indicators more accurate in the experiment and have more reference value.In addition,the fusion image quality index based on gradient feature is analyzed in detail,and the relationship between gradient direction change and image quality change is studied,which proves the rationality of gradient direction similarity measurement.(3)In this thesis,an image fusion index algorithm based on structural similarity is proposed,which combines the characteristics of human visual system.The requirement of fusion application scenarios is analyzed,and the quality index algorithm is modified for infrared and visible image fusion scenarios.(4)In this thesis,an experiment based on multi-focus image fusion is designed to verify the rationality of image fusion quality index from an objective point of view and solve the problem of lack of absolute truth value in image fusion index verification.