Parameter estimation in the general contourlet pansharpening method using Bayesian inference

This paper solves the problem of parameter estimation for the general contourlet pansharpening method using Bayesian inference. In the general contourlet panshapening method, a set of parameters that control the contribution of each band of the multispectral image, the panchromatic image and the prior knowledge on the image need to be set. The proposed method takes into account the relationship between contourlet coefficients to incorporate prior knowledge on the unknown parameters in the form of hyperprior distributions. This method is able to estimate all the unknown parameters together with the high resolution multispectral image in a fully automatic way. The experimental results show that the proposed method not only enhances the spatial resolution of the pansharpened image, but also preserves the spectral information of the original multispectral image.