An exposure fusion approach without ghost for dynamic scenes

Abstract

In this paper, we present a new method for multi-exposure images fusion of dynamic scenes to remove ghost artefacts. First we unify the exposure level of each input image to that of the reference image and generate ghost detection binary images using difference method with an adaptive threshold strategy. Then we modify the binary images further with morphological operations. Finally, we redefine weight maps of exposure fusion according to the modified binary images to obtain the fused image. We can generate different deghosting result base on an optional reference image. Our method achieves high-quality deghosting effect even if the reference image is not well exposed. Besides, our method can process the misaligned exposures caused by camera shake effectively. Experimental results of various dynamic scenes prove the improvement over previous reference-based deghosting methods.