An Edge Preserving Multiresolution Fusion Using Curvelet Transform and Mrf Prior

Satellites used for remote sensing are, in general, capable of acquiring two different types of images: multispectral (MS) images and a panchromatic (Pan) image. The MS sensors provide multiband images with accurate spectral (color) information but with low spatial resolution. In this project, we propose an edge preserving multiresolution fusion by using subsampled as well as nonsubsampled CT. a close approximation (initial estimate) to the final fused image is first obtained using the available Pan, MS image, and the CT/NSCT, which is then used in deriving the relation between LR and HR MS images and also the edges in the final fused image. The final fused MS image is obtained by using regularization for which we use a homogenous MRF prior that requires simple gradient based optimization in order to obtain the final solution.