Spectral and spatial super-resolution method for Earth remote sensing image fusion

In the article we propose a spatial and spectral super-resolution algorithm for a set of multichannel images obtained by various Earth remote sensing detectors. We regard the result of the algorithm as a model of an ideal data source, which has a better accuracy of the observed terrain representation than each of the input images having lower spatial and spectral resolution. The proposed algorithm utilizes a method of gradient descent and applies a refined model of image observation, including both spectral and spatial down-sampling and up-sampling. The article describes an experimental study of the proposed algorithm and a comparison of the quality of its work with bilinear interpolation of low-resolution images. The practical application of the proposed algorithm consists in the joint processing of remote sensing data of various levels, which makes it possible to erase the boundaries that arise from the design differences of imaging sensors.