Pyramidal dual-tree directional filter bank based exposure fusion for two complementary images

Currently, most exposure fusion methods are easy to be influenced by the location of object in the image. In order to reduce the influence caused by un-alignment in the source images, shift-invariance multiscale decomposition can be used to fuse images. However, commonly used shift-invariant decomposition methods have high redundant ratio, which limits their application. So, redundant ratio of decomposition is an important factor, which must be considered for exposure fusion method. In this paper, a novel exposure fusion method is proposed, which is based on a pyramidal dual-tree directional filter bank (PDTDFB) having both shift-invariance and low redundant ratio characteristics. Furthermore, different fusion rules are used for coefficient selection in the lowpass and highpass subimages. Experiments show that the proposed method can extract more important visual information from input images than typical exposure fusion methods.