Hyperspectral image super-resolution reconstruction method based on depth image prior

Abstract

According to the hyperspectral image super-resolution reconstruction method based on depth image prior provided by the invention, an automatic processing network not based on training is added in an existing subject learning network, and the input processing module is designed to fully adjust network input according to a network subject structure, so that the input and the network structure generate resonance, and the super-resolution reconstruction of the hyperspectral image is realized. Therefore, the prior information of the image is fully utilized, the image is firstly captured by utilizing the intrinsic characteristics of the DCNN, and then the image is recovered. The network structure provided by the invention makes full use of the correlation between the spatial information of the hyperspectral image and the spectral band to learn the image features, and the input processing module in the automatic processing network can automatically adjust the input structure, so that the application of the DIP algorithm can be greatly expanded, the DIP method is more suitable for the hyperspectral super-resolution task, and the efficiency of the hyperspectral super-resolution task is improved. And the quality of the composition image can be further improved.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence and image processing, and in particular relates to a hyperspectral image super-resolution reconstruction method based on depth image prior. Background technique [0002] Hyperspectral images have rich spectral and spatial information and have advantages in capturing the details of target objects. They have a very wide range of applications, such as land, atmosphere, ocean and other fields of observation. However, due to the limitation of incident energy, in the design of optical remote sensing system, if other factors remain constant and high signal-to-noise ratio (SNR) is guaranteed, the spatial resolution decreases with the increase of spectral features. Hyperspectral image super-resolution reconstruction technology aims to reconstruct a high-resolution hyperspectral image from a low-resolution hyperspectral image. [0003] The existing technology proposes hyperspectral image s...

AI Technical Summary

Problems solved by technology

However, the number of available hyperspectral image datasets is small compared to natural images

2) Considering the different imaging conditions and the number of hyperspectral bands, it is still a difficult problem to construct a unified deep neural network

[0006] 1) The proposed structure is based on the traditional grayscale / RGB image super-resolution task, so the method cannot achieve satisfactory results when dealing with high-dimensional hyperspectral images;

Images