Satellite remote sensing image super resolution based on markov random fields

This paper studies satellite remote sensing image super resolution that employs image processing techniques to reconstruct the high-resolution image from a set of low-resolution observations of the same scene. A heuristic approach for maximum a posteriori (MAP) estimate of desired high-resolution image based on markov random fields (MRF) is presented. Under the posteriori distribution deduced by Bayesian criterion, the reconstruction image is derived by finding the global optimized estimation with the simulated annealing (SA) optimization mechanism. In the experiments, the proposed method is evaluated in a simulated framework that the estimate images are compared with the reference one using Normalized Mean Square Error (NMSE) criterion. The results quantitatively indicate the super performance of super resolution reconstruction and noise robustness obtained by our approach in comparison with the Cubic interpolation.