Method and apparatus for super-resolution reconstruction of remote sensing images based on mixed random downsampling

Abstract

A method and apparatus for super-resolution reconstruction of remote sensing images based on mix random downsampling are provided in that embodiments of the present invention. The method comprises thefollowing steps: the multi-scale metre-resolution optical remote sensing images of the target to be reconstructed are mixed and randomly downsampled to obtain a high-resolution image matrix and a low-resolution image matrix, and a high-resolution image block mapping matrix and a low-resolution image block mapping matrix are constructed; A super-resolution reconstruction model for optical remote sensing image super-resolution reconstruction is obtained by using that high-resolution and low-resolution image block mapping matrix as a training sample and carrying out depth learning training; A superresolution reconstruction is performed on an optical remote sensing image of an object to be reconstruct based on that superresolution reconstruction model. Considering the optical imaging characteristics of satellite sensor load, a hybrid stochastic desampling model is used to generate low-resolution image of optical sensor, which enhances the scene adaptability of super-resolution reconstruction model, processing accuracy and reliability of the results.

Description

technical field [0001] The present invention relates to the technical field of image processing, and more specifically, to a remote sensing image super-resolution reconstruction method and device based on hybrid random downsampling. Background technique [0002] Image super-resolution reconstruction technology (Super-Resolution) is an important digital image processing technology, which uses one or more low-resolution images (or motion sequences) to reconstruct a high-resolution, Images with high information content. This technology breaks through the resolution limitation of the image sensor itself, and can increase the resolution of the image and improve the image quality without changing or improving the image acquisition hardware. The processing results are conducive to surface visual interpretation, image target algorithm recognition, surface object type analysis, quantitative inversion accuracy improvement, and improve the information expression ability and utilizatio...

AI Technical Summary

Problems solved by technology

Especially for dynamic scenes, there are moving objects in the scene, and motion estimation is very difficult

[0004] At present, artificial intelligence technology based on deep learning has been maturely applied in image target super-resolution reconstruction, such as ESPCN, SRCNN, SRGAN, VDSR, FSRCNN and other image super-resolution reconstruction technologies based on convolutional neural network technology, but the existing The preprocessing method involved in the sample construction stage of the deep learning method is too simple, and its adaptability and accuracy to satellite remote sensing image processing are poor. It is difficult to combine with a high-quality neural network model to obtain more significant processing effects and reduce overresolution. Practical Application Value of Reconstruction Technology

Images