Research On Image Enhancement Algorithm Based On Variational Framework

In recent years,with the rapid development of computer vision technology,people have higher and higher requirements for the quality of video and images,but due to the limits of the shooting environment and shooting equipment,etc,the collected images are often prone to problems such as difficulty in identifying objects,unclear details,color distortion,and so on.Image enhancement refers to highlighting certain information in an image according to specific needs,while weakening or removing some unwanted information.The purpose is to improve the overall quality of the image.It has important application value,so it has always been a research hotspot for scholars.This paper mainly proposes two image enhancement algorithms under the variational framework and based on Retinex theory and designs an image enhancement prototype system.1.In view of the problem of unclear recognition of low illumination image information and fuzzy display of details,this thesis proposes an illumination compensation and detail adjustment algorithm based on variational framework.Firstly,based on Retinex theory,a reasonable objective function is used in the variational framework.The illumination component and the reflection component of the image are solved by the alternate solution method,and the morphological combination is used to obtain the first reference map with the variation enhancement.Secondly,the illumination compensation and detail adjustment are performed on the reference image to obtain the other two.In the end,the three benchmarks are subjected to Laplacian fusion,and the fusion is performed to restore the details to obtain the final enhanced image.By conducting experiments in public data sets,the algorithm can achieve good results both subjectively and objectively.2.Aiming at the problems of color distortion and overexposure for low illumination image fusion,this thesis proposes a multi-scale fusion algorithm for low illumination image based on variational framework.Firstly,three reference maps are obtained by the method of illumination compensation and detail adjustment based on the variational framework.Secondly,the weighted values of the three reference images are designed based on the three dimensions of brightness,exposure and hue of the image.The pixels are assigned a larger weight,and the underexposed pixels are assigned a smaller weight to obtain an exposed image set.Finally,the exposed image set and the reference image setare combined by multi-scale fusion to output the enhanced image.The algorithm is tested on different data sets,and the subjective analysis and objective conclusions are better than other comparison algorithms.3.This thesis designs and develops an image enhancement prototype system based on variational framework.The system combines the image enhancement algorithm proposed in this thesis,which can effectively realize the function of one-click enhancement of the target image.