Research On Fusion Algorithm Of Multi-exposure Sequence Image

With the continuous innovation of science and technology,people have higher and higher requirements for image quality.However.The images captured by ordinary cameras often have problems such as color distortion and lack of details,which cannot restore the real scene that human eyes can see.In order to overcome the shortcomings of ordinary cameras,high dynamic range imaging technology has emerged.The high dynamic range images generated through high dynamic range imaging technology can better restore real scenes,greatly promoting the development process of digital information.Among the methods to generate high dynamic range images,multi-exposure image fusion technology has become one of the research hotspots for scholars at home and abroad.This technology utilizes three or more images with different exposure levels to fuse into a High Dynamic Range(HDR)image with rich colors and details.According to different shooting scenes,it can be divided into two categories: static scenes and dynamic scenes.This paper conducts in-depth research on the lack of detail in the bright and dark regions of fusion images in static scenes and the ghosts of fusion images in dynamic scenes.Based on the current traditional multi-exposure image fusion methods,further improvements have been made.The main research contents of this paper are as follows:1.Aiming at the lack of detail in the bright and dark areas of images in the static scene fusion method,this paper improves the multi-scale fusion algorithm based on the traditional Laplacian pyramid.When constructing the weight map,two indicators,exposure and color saturation,were selected,and the Weighted Guided Image Filter(WGIF)algorithm was used to filter the pyramid decomposed weight map to reduce halo artifacts.After obtaining the basic fusion image,select the bright and dark regions of the LDR(Low Dynamic Range)image with a specific threshold and assign high weights to the corresponding bright and dark regions of the basic fusion HDR image for weighted fusion,so as to restore the details of the bright and dark regions to the maximum extent.2.Aiming at the ghost problem caused by multi-exposure image fusion in dynamic scenes,this paper sets up a method of motion region detection and applies the Image Poisson Editing algorithm to eliminate ghosting.This method needs to detect the moving target completely,but because of the different exposure levels of sequence images,the difficulty of detecting the moving target increases.Therefore,this method needs to first perform histogram matching processing on adjacent images to obtain image pairs with basically the same exposure,and then use difference method,maximum inter-class variance method and morphological processing to detect the initial moving target.In order to completely extract the moving target,further find out the accurate edge according to the detected initial moving target edge.Finally,the motion region of an image with good exposure is fused into the basic HDR image using the Image Poisson Editing algorithm to obtain a HDR image without ghosts.