Earth remote sensing data (ERS) obtained by various sensors are usually different in spatial and spectral resolution, and in the registration time as well. Changes in observation conditions lead to brightness distortions, some of which are compensated by atmospheric correction. Unfortunately, the distortions in the composition of the scene (scene distortions) cannot be compensated by means of atmospheric correction and requires fusion methods to be applied. The existing super-resolution methods for remote sensing data are based on the assumption that there are no any scene distortions in the analyzed images. In this article we propose an algorithm for combining ERS data with an increase in spectral and spatial resolution taking into account scene distortions. We found that the use of a larger dataset including images with scene distortions reduces root restoration error of 2-4% in average if the dataset contains a small number of images without scene distortions (from 2 to 6). With the larger number of images without scene distortions, the error value decreases by 1%. If all the images in the input dataset contain scene distortions, the proposed algorithm achieves super-resolution restoration of the scene image as well.
