Research On Adaptive Complementary Exposure And Fusion Algorithm For High Dynamic Images

The acquisition and display of high dynamic images and videos has been a hot issue in the field of multimedia processing in recent years.With the rapid upgrading of computing and display technology,people have gradually increased the requirements for video fidelity and detail restoration quality.High dynamic technology is the key technology for restoring natural scenes,so it has attracted more and more attention from researchers.High dynamic imaging technology based on multi-exposure image fusion is an effective method to present images of high dynamic scenes under conditions where the physical characteristics of the camera and display devices are limited.However,current fusion algorithms need to take multiple images with increasing exposure time as input sources,and calculate a high dynamic image as output by fusing pixel values of different images in the same area.This method has inherent flaws.On the one hand,the preset exposure parameters cannot adapt to the changing dynamic range of the incident light.On the other hand,the fusion of multiple images brings a lot of computational load.These defects make this method only suitable for the generation of high dynamic images in static scenes,and cannot adapt to the capture of high dynamic videos.For this problem,this thesis proposes an adaptive complementary exposure algorithm that can automatically adjust two complementary exposure parameters and capture two images with minimal redundancy as the basis for high dynamic images synthesis.At the same time,this thesis proposes a complementary exposure image fusion in dynamic scenes.This algorithm can be applied to the generation of highly dynamic video.The specific research content of the thesis is as follows:(1)For the problem that the existing high dynamic image data set has only a small number of images in the same scene and cannot fully verify the high dynamic algorithm,this thesis builds a high dynamic database by capturing four sets of fine step image data sets.The database fully simulates the image characteristics in high dynamic scenes,and provides a basis for the analysis and design of high dynamic algorithms.(2)Fot the problem that the exposure parameters in the existing algorithms do not adapt to high dynamic scenes,this thesis proposes an adaptive complementary exposure algorithm to determine the exposure parameters of different high dynamic scenes.The algorithm achieves the goal of automatically capturing high dynamic images by improving the exposure strategy in the ISP without replacing the existing sensors.The algorithm is verified through the constructed high dynamic database and the existing high dynamic data set.Experimental results show that the algorithm can automatically adjust two complementary exposure parameters to capture two images with minimal redundancy.In addition,the algorithm is transplanted to the camera platform,and experimental evaluation shows that the algorithm can meet the real-time requirements.(3)For the ghost image generated by the existing image fusion algorithms for complementary exposure images,this thesis proposes a complementary exposure image fusion algorithm.This algorithm proposes a ghost region detection algorithm based on numerical analysis,and uses a joint histogram-based algorithm to compensate for ghost regions.The experimental evaluation of the algorithm on the complementary exposure scene dataset shows that the fusion image obtained by the algorithm in this paper has no ghost images and richer texture details.In addition,compared with the existing algorithms,the algorithm in this thesis has obvious advantages in execution time,and has achieved higher values under the evaluation indicators of PSNR and SSIM.