Unsupervised remote sensing image super-resolution reconstruction method based on image recursion

Abstract

The invention discloses an unsupervised remote sensing image super-resolution reconstruction method based on image recursion. Obtaining an original low-resolution image; extracting a degradation kernel from the original low-resolution image; performing down-sampling on the original low-resolution image by using a degradation kernel to obtain a low-resolution image after down-sampling; training the super-resolution reconstruction network model based on the original low-resolution image and the down-sampled low-resolution image; and adopting the super-resolution reconstruction network model to test the to-be-tested low-resolution image to obtain a corresponding super-resolution reconstruction result. Through the method, a super-resolution reconstruction method can be applied to the remote sensing image, so that the accuracy of obtaining pairing data from the remote sensing image is improved, and a network realizes a good reconstruction effect on downsampling used for reconstruction synthesis.

Description

technical field [0001] The invention belongs to the technical field of digital image processing, in particular to an unsupervised remote sensing image super-resolution reconstruction method based on image recursion. Background technique [0002] Super-resolution reconstruction is a method of reconstructing low-resolution images with less details into high-resolution images with rich texture details. Single-frame image super-resolution reconstruction means that the input image data is non-sequential images, and the single-frame image Image processing tasks that process low-resolution image data into high-resolution images with rich texture details and better visual effects. At present, the super-resolution reconstruction of remote sensing images mostly adopts deep learning methods, and the supervised image super-resolution reconstruction methods based on deep learning are all trained by image pairs of low-resolution images and high-resolution images of the same area. It is v...

AI Technical Summary

Problems solved by technology

It is very difficult to obtain paired remote sensing image datasets because it is impossible for the imaging sensor on the satellite to simultaneously obtain image pairs of low-resolution images and high-resolution images of the same area. Some supervised image super-resolution reconstruction methods use synthetic The image pair for training, the synthesized image pair is generated using a fixed downsampling method, such as bicubic downsampling, the degradation method of the actual low-resolution remote sensing image may be very different from the fixed downsampling method, which also leads to the training model Poor performance in reconstructing real low-resolution remote sensing images

Images