The Algorithm Of Exposure Image Fusion Based On Multiresolution Analysis

With the rapid development of computer vision,the requirement of the people for images is getting higher and higher.High dynamic range imaging technology has improved the quality of digital images greatly and has become one of the research hotspots in the field of digital images.Multi-exposure image fusion is an effective method for obtaining high dynamic range images.It can expand the dynamic range of the image and enhance the detail of the image by blending a series of low dynamic range images of different exposures.Multi-exposure fusion has significant value for research and applications,and has received wide attention.Multi-exposure image fusion can be divided into static scene fusion and dynamic scene fusion according to the shooting scene.The main research of multi-exposure fusion in static scenes is how to preserve more details as much as possible on the basis of minimum visual distortion.The main research of dynamic scene multi-exposure fusion is how to remove the ghosting problem caused by moving objects.The related techniques and methods of multi-exposure image fusion are studied in this thesis.Aiming at the loss of details in static scene,an improved multi-exposure fusion with detail enhancement based on multi-resolution analysis image fusion and a multi-exposure fusion based on multi-scale detail enhancement were proposed.Then,for the multi-exposure fusion in the dynamic scene,an algorithm with ghosting removal based on superpixel segmentation was proposed,which can detect and remove ghosting effectively.The main work of this thesis includes:1.The current research of multi-exposure fusion was described,and the image registration for exposures,multi-exposure fusion methods for static scenes and dynamic scenes were studied and analyzed in detail.2.For the loss of details in multi-exposure fusion of static scenes,two strategies for detail enhancement based on multi-resolution analysis were proposed.Multi-resolution image fusion can avoid the generation of seams and reduce visual distortion effectively,but the traditional multi-resolution image fusion will lead to the loss of details.An improved multi-exposure fusion with detail enhancement based on multi-resolution analysis image fusion and a multi-exposure fusion based on multi-scale detail enhancement were proposed in this thesis.The former enhances the high frequency information of the image pyramid by using the exposure brightness of each image as a detail enhancement factor.The latter extracts the multi-scale detail maps of input images,and then enhances the fused image directly.3.For dynamic scenes,moving objects will lead to ghosting.An algorithm with motion detection and ghosting removal based on superpixel segmentation was proposed in this thesis.First the input sequence is adjusted to the same exposure level according to the reference image,and the image difference is calculated.Then,the contour is extracted by the SLIC,and the moving object region is further detected according to the proportion of the moving pixels in each superpixel region.Finally,the fusion weight of the motion is reduced to obtain a fused image without ghosting.