Research On The Multi Exposure Image Fusion And Quality Assessment

Image fusion technology refers to the process of the information of different images under the same target to maximize the available information of each image and to skillfully blend into a high-quality image.With the development of the technology,image fusion has produced many topics such as multi-view image fusion,multi-exposure image fusion,and multi-focal image fusion.As an important direction of image fusion technology,multi-exposure image fusion has drawn much attention from researchers in recent years.Multi-exposure image fusion algorithm is to fuse multiple exposure images taken in the same perspective.Due to the limitation of the light of shooting scene or shooting equipment,it may be difficult to express all details in a single image very clearly at some times.Consequently,it may cause serious overexposure or underexposure.Multi-exposure fusion algorithm can compare the intensity,contrast ratio and other information in different regions of the image and fuse the useful information of the image to obtain a high-quality result image.In this paper,we focus on the problem of loss of detail information and inconsistency of regional brightness in the block-based multi-exposure image fusion algorithm.Firstly,the pixels are clustered according to the intensities of the pixels in the multi-exposure image group,and the images are divided into different image blocks.Then according to the image contrast ratio,color saturation and exposure make the best exposure selection for each image block.After that,the contribution degree is adjusted by the intensity information between the image blocks to ensure that the brightness of the fused image is consistent.Finally,the images are fused together smoothly using a smoothing filter that retains the details.At the same time,this paper presents a multi-exposure fusion image quality evaluation algorithm.The traditional multi-exposure fusion quality evaluation algorithms mainly include the evaluation algorithm based on statistical properties,the evaluation algorithm based on information volume,and the evaluation algorithm based on human vision.But all of them have very strong limitations.To solve the problem,this paper proposes a multi-exposure image fusion algorithm based on the detail preservation and regional brightness consistency.And the structure preserving algorithm mainly compares whether the information expressed in the image before and after fusion is consistent;the natural image statistics algorithm mainly compares whether the fused image is natural or not.Finally,the two algorithms are combined to propose the quality evaluation algorithm based on the structure preserving degree and the natural image statistics described in this paper.Finally,the experiments based on the Daniel Scharstein database,Keda Ma and the self-photographing multi-exposure image group all achieved very good results.At last,the present work is summarized and the future prospects are prospected.