Fusion of Panchromatic and Multispectral Images Using Multiscale Convolution Sparse Decomposition

In this article, we proposed a novel image fusion method based on multiscale convolution sparse decomposition (MCSD). A unified framework MCSD is first utilized to decompose panchromatic (PAN) and the spatial component of upsampled low resolution multispectral (LR MS) images, which can produce corresponding frequencies feature maps. By combining with analysis, efficiently approximate spectral information in images. Next, binary map generated from gradient integrate LR MS PAN For maps, gain for each pixel calculated according similarity between local patches them. Finally, fused reconstructed by sum frequency Some experiments are conducted QuickBird GeoEye-1 satellite datasets. Compared other methods, performs better visual numerical evaluations.