Hyperspectral image super-resolution method based on spectral constraint adversarial network

Abstract

The invention provides a hyperspectral image super-resolution method based on a spectral constraint adversarial network. The problems that a hyperspectral image generated in the prior art is serious in spectral distortion, depends on prior information and is high in operation complexity are solved. The method comprises the following steps: constructing a generator network; constructing a decisiondevice network; constructing a spectral constraint adversarial network; initializing a spectral constraint adversarial network; generating a training set; training a spectral constraint adversarial network; and performing super-resolution on the hyperspectral image. According to the method, the spectral constraint adversarial network is utilized, high-resolution multispectral images in the same scene do not need to be used as prior information for image fusion, the spatial resolution of the hyperspectral images can be effectively improved, and meanwhile spectral distortion of the hyperspectralimages after super-resolution is reduced.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to a hyperspectral image super-resolution method based on a spectral constraint confrontation network in the technical field of image super-resolution. The invention can be used to improve the spatial resolution of hyperspectral images with relatively low spatial resolution. Background technique [0002] Since the hyperspectral image contains both the spatial information and the spectral information of the scene being shot, compared with the single way that the natural image can only be processed in the spatial dimension, the hyperspectral image can be processed from the spatial dimension and the spectral dimension at the same time. In natural image processing fields such as target detection, image classification, and semantic segmentation, higher spatial resolution often means better results can be obtained, and the same is true for hyperspectral images. However, th...

AI Technical Summary

Problems solved by technology

Although this method takes into account the use of generative confrontation network for hyperspectral image super-resolution, it solves the problems of insufficient spatial resolution, low definition and poor visual effect of existing hyperspectral images.

However, the shortcomings of this method are that the network structure for extracting the spectral information of the hyperspectral image and the spectral constraints of the set loss function are insufficient, which makes the spectral distortion of the high-resolution hyperspectral image reconstructed by this method serious.

Although this method considers the use of fusion method to solve the problem of low spatial resolution of the enhanced hyperspectral image produced by the approximate Heaviside function sparse representation method, the disadvantage of this method is that the approximate Heaviside function used by this method The sparse representation method of Weiside function has more nonlinear operations, which makes the method more complex

Images