Multisensor image fusion based on biorthogonal multiwavelet transform and region competition

This paper proposed a novel region based novel multisensor image fusion algorithm. A graph based segmentation method and biorthogonal multiwavelet transform (BMWT) is combined to segment the features of the source images to produce a region map. Then we use a scheme, which calculates the characteristics of each region of sources in wavelet domain by averaging and selection, to fuse the images. The experiment results demonstrate that our proposed algorithm is comparable to pixel based fusion methods, moreover has many advantages such as reduced sensitivity to noise and mis-registration, more flexible fusion scheme despite its increase in complexity.