Multi-exposure Image Fusion Based On Guided Filtering And Ghost Removal

With the development of digital image,the demand for image quality is increasing.The emergence of high dynamic range imaging technology meets people's needs.High dynamic range imaging technology restores the light and the dark information of the actual scene,and this technology is more in line with people's visual perception.High dynamic range imaging technology promotes the development of digital information.Therefore,high dynamic range imaging technology has wide application prospects.At present,the more common method for high dynamic imaging is multi-exposure image fusion.Multi-exposure image fusion realizes high dynamic range imaging by merging low dynamic range image sequences with different exposures in the same scene.The fusion algorithm of static scene and the ghost removal algorithm of dynamic scene are studies in this thesis.Aiming at the fusion algorithm of static scene,a multi-exposure fusion algorithm based on optimized guided filtering is proposed.The algorithm effectively removes halo and presents rich details.Aiming at the ghost removal algorithm of dynamic scene,a ghost detection and removal algorithm based on the reference image are proposed.The algorithm accurately detects moving objects.The research contents of this thesis include:1.The detailed information and brightness information of the fusion image are analyzed in depth,and the local contrast and proper exposure of the image sequence are redesigned.Local contrast is calculated by combining phase consistency and edge detection operators to measure structural information and contrast changes in local areas.Proper exposure is added by a balance factor that balances global contrast and local exposure.According to the quality measurement factor,the initial weight of the image is evaluated.And the effective information of each image is accurately extracted.2.Aiming at the problems of halo,light and dark flip of the multi-exposure fusion algorithm,a multi-exposure fusion algorithm with optimized guided filtering is proposed in this thesis.In view of the advantages of the guided filtering,the noise of the basic weight is eliminated.According to the disadvantages of the guided filtering,an improved strategy is proposed for the guided filtering.The optimized guided filtering adapts to regional texture differences,identifies and strengthens the edge regions,and prevents blurring at the edges.3.Aiming at the problem of inadequate ghost removal algorithm,a new ghost image removal strategy based on reference images is proposed in this thesis.Firstly,the reference image and the non-reference image are matched to eliminate the influence of the exposure difference on the detection of moving pixels.Secondly,the approximate contour of the moving object is determined by the difference method.Thirdly,in order to ensure that similarly structured pixels have similar fusion contributions,the maximum inter-class variance is used to further detect moving pixels.Finally,the inconsistency caused by the relative motion between the reference image and the non-reference image is eliminated.And the motion bitmap is corrected.