Super-resolution Inversion and Reconstruction for Remote Sensing Images of Bands of Interest

Abstract

The super-resolution inversion and reconstruction algorithm of remote sensing image of band of interest is proposed in this paper. Based on the existing multi-resolution hyperspectral images, a spectral reflectance dictionary is constructed, according to which the objects in radiation calibrated and atmospheric corrected scene spectral reflectance images are sorted and classified using the end element extraction algorithm and the correlation distance method. Based on the classification results, a dictionary of scene feature spectral reflectance is constructed. Based on this, the sparse representation method and the secondary optimization algorithm based on image similarity are used to solve and optimize the feature distribution of the scene image. According to the feature distribution of the scene image, using the remote sensing link imaging model, super-resolution inversion reconstruction is performed to obtain a high-resolution remote sensing image of a band of interest. Experimental results show that the proposed algorithm can reconstruct low-resolution remote-sensing images of different bands into high-resolution remote-sensing images. Thus, the proposed algorithm can effectively boost image resolution, enrich image details and improve the target detection capability by the image.