Research On Multi-exposure Fusion High Dynamic Range Image

With the continuous innovation of science and technology,the field of computer vision and image processing is developing rapidly,and people’s requirements for the quality of digital images are getting higher and higher.However,because ordinary digital cameras and other imaging devices compress the dynamic range of brightness of real scenes,the captured images have halo,gradient inversion,color distortion,etc.,which makes it impossible to truly restore the scene information that the human eye can see.The high dynamic range image makes up for the above imaging deficiencies by expanding the dynamic range of the image.And the best method is the multi-exposure image fusion method.Multi-exposure image fusion technology is the use of three or more images of the same scene with different exposure,in the image transformation domain or spatial domain to carry out some image processing operations,fusion into a high-definition,color details are rich result image.In this paper,two algorithms are proposed,namely,multi-exposure image fusion algorithm based on improved pyramid and multi-exposure image fusion algorithm based on recursive filter.These two algorithms improve the image quality to a certain extent,improve the halo,color distortion and other problems,and achieve a good fusion effect.The main research contents of this paper are as follows:(1)Aiming at the uneven distribution of fused image detail information,this paper proposes an improved pyramid algorithm.First,perform adaptive histogram equalization processing on the original image sequence to improve the edge texture information of the image;secondly,perform a weighted average of the feature factors(contrast factor,good exposure factor,entropy factor)to obtain the fusion weight;build the weight pyramid,It is fused with the original image pyramid obtained after the Laplace transform,and finally the fused image is generated through the inverse Laplace pyramid transform.(2)In this paper,a recursive filter based multi-exposure fusion method is proposed to solve the problems of halo and gradient flip in the fusion image.The flow of the algorithm is to perform recursive filtering operations on the original image sequence to obtain the base layer of the image,and then subtract the original image from the base layer image to obtain the detail layer image.Then separate the base layer and the detail layer of the image for different operations.The base layer uses three quality metrics(local contrast,brightness,information entropy)to extract image detail features,calculates the initial fusion weight of the base layer,and then uses a recursive filter to modify and refine the initial weight;the detail layer uses weighted average The method to obtain the detail layer to be merged.Finally,the corrected weight map is fused with the processed detail layer to obtain the final fused image.