Remote sensing image super-resolution reconstruction method based on multi-dictionary learning and non-local information fusion

Abstract

The invention provides a remote sensing image super-resolution reconstruction method based on multi-dictionary learning and non-local information fusion. The method provided by the invention takes the non-local similarity of remote sensing images into consideration on the basis of image reconstruction by using learning dictionaries, information contained in a low-resolution image is added to a reconstructed image, and high-frequency details are enabled to be richer in content. Meanwhile, images reconstructed by multiple dictionaries are fused through a principle of low-rank decomposition, thereby effectively utilizing non-redundant information contained among the plurality of images. Finally, the algorithm further reduces a reconstruction error of the images by using global optimization.

Description

technical field [0001] The invention relates to the field of image processing and remote sensing information intelligent processing, in particular to a remote sensing image super-resolution reconstruction method based on multi-dictionary learning and non-local information fusion. Background technique [0002] With the continuous development of remote sensing technology, the spatial resolution of satellite remote sensing images is getting higher and higher, but satellite remote sensors still face high input costs and limitations of hardware production technology. Therefore, improving image resolution from the aspects of algorithms and software has become a hot research topic at present. The resulting image super-resolution technology can use low-resolution images to reconstruct high-resolution images, thereby overcoming the inherent resolution limitation of imaging equipment, and has important application value in the field of satellite remote sensing imaging. [0003] At pr...

AI Technical Summary

Problems solved by technology

In real situations, the imaging conditions of remote sensing images are complex, the degradation model is uncertain, and the types of ground objects are also the same. The image quality of super-resolution reconstruction using a single dictionary is sometimes even inferior to traditional interpolation algorithms

[0007] 2. Only external information is considered, that is, the training dictionary is used for image reconstruction, and the structural information contained in the image itself is ignored

Remote sensing images have high texture complexity and strong non-local self-similarity

This special structure, which is different from natural images, often results in a large amount of non-local self-similarity information in the same remote sensing image, such as roads, farmland, houses, etc., which cannot be effectively utilized by traditional methods.

Images