A Spectral Reconstruction Method Based on Few-Band High-Resolution Images

Abstract

Provided is a high-quality spectrum reconstruction method based on a few-wave-band high-resolution image and a low-resolution hyperspectral image. The method comprises the steps that the high-resolution image having few wave bands and the low-resolution hyperspectral image under the same scene are obtained; the low-resolution hyperspectral image is trained in a non-decomposition mode to obtain a spectrum dictionary; then, the high-resolution image having the few wave bands is subjected to sparse representation under the condition of being free of non-negative restriction, and a sparse representation coefficient is obtained; the part which cannot be represented by a sparse representation framework is estimated through spatial structure information; the dictionary, the sparse coefficient and the estimation part are utilized for reconstructing the hyperspectral image with high resolution. Accordingly, under the sparse representation framework, the non-decomposition mode is introduced to solve the spectrum dictionary, the deficiency of descriptive power of endmember decomposition to spectral characteristics of the hyperspectral image is made up, and the reconstructed spectrum precision, the effectiveness of hyperspectral image reconstruction and the spatial accuracy of reconstructing the hyperspectral image are effectively improved.

Description

technical field [0001] The invention belongs to the field of image processing, is suitable for hyperspectral remote sensing image reconstruction, and in particular relates to a high-quality spectral reconstruction method based on a few band high-resolution images and low-resolution hyperspectral images. Background technique [0002] Hyperspectral images are composed of a large number of single-band images, and each pixel in the image has a quasi-continuous spectral curve. In the hyperspectral imaging process, due to the narrow spectral bandwidth, a large instantaneous field of view (IFOR) must be used to accumulate enough light quanta to maintain the signal-to-noise ratio of imaging, and the increase of the instantaneous field of view will reduce the resolution of the image. . However, in many application fields of hyperspectral images, such as object recognition and classification, and environment detection, high-resolution images are indispensable, so it is of great signi...

AI Technical Summary

Problems solved by technology

However, there are certain difficulties in the number of endmembers and the extraction of pure endmembers in the decomposition of mixed pixels, so the obtained hyperspectral image has spectral distortion, and it is impossible to obtain satisfactory results in high-resolution hyperspectral image reconstruction. result

Images