The new segmentation and fuzzy logic based multi-sensor image fusion

In this paper, we propose a new multi-resolution fusion algorithm for spatially registered multi-sensor images fusion. First, we use a modified watershed algorithm to produce a region map of source images. This segmentation is then used to generate region-based decision map, which is obtained based on local texture features in the dual-tree discrete wavelet transform domain. The region-based decision map is very accurate for selecting high frequency coefficients between the source images compare to the pixel-based decision map, but very important details of source images is missed in the final fused image. In this study, we propose an algorithm based on fuzzy logic, which uses both pixel and region-based decision map based on a dissimilarity measure of source images for fusion rule. Also a region-based activity measure of each source image is used for low frequency fusion rule instead of simple averaging. Our new method provides improved subjective and objectives results compared to the previously pixel and region-based fusion methods.