A New Variational Model for Panchromatic and Multispectral Image Fusion

Image fusion is a powerful tool to provide a single image which contains the complementary information from the input images. A novel variational model is presented for panchromatic and multispectral image fusion. Based on the Socolinsky's contrast model, an advanced energy minimization problem is posed to find the solution whose gradient is closest to that of the panchromatic image. To improve the multispectral image's spatial resolution and preserve its spectral information as much as possible, spectral coherence, interband correlation and contrast enhancement terms are explicitly enforced into the fusion process. The performance of our model is evaluated on several IKONOS and QuickBird datasets. Experimental results show that our model can produce images with both high spatial and high spectral quality.