A hyperspectral image denoising method and device

Abstract

The invention provides a hyperspectral image denoising method and device. The method includes: S1, obtaining at least one effective band that is similar to the structural information of the noise band of the hyperspectral image and has better image quality than the noise band; S2, An objective function is established based on the at least one effective band, and a denoised clear band is obtained based on the objective function. By utilizing the correlation between the bands, the denoising problem of the noise band is transformed into the fusion of the bands, and the clear band after final denoising is obtained. It considers and overcomes the neglect in the previous methods that the noise may only exist in some or certain areas. band situation and achieved better denoising results.

Description

technical field [0001] The present invention relates to the field of hyperspectral image denoising, and more specifically, to a hyperspectral image denoising method and device. Background technique [0002] Compared with RGB color or grayscale images, hyperspectral images have richer band information, which can improve the efficiency of many computer vision applications, such as face recognition, remote sensing, and motion segmentation. However, in practice, hyperspectral images will inevitably be affected by noise, such as sensor defects or the environment, so hyperspectral denoising is very meaningful research. The formation process of the noise image can be simply expressed as y=x+n, where n is the noise, x is the clear original image, and y is the noise image. However, the existing hyperspectral image denoising methods seldom consider a practical problem: the noise in the hyperspectral image does not necessarily exist in all bands, but may only exist in certain or speci...

AI Technical Summary

Problems solved by technology

However, the existing hyperspectral image denoising methods seldom consider a practical problem: the noise in the hyperspectral image does not necessarily exist in all bands, but may only exist in certain or specific bands. These hyperspectral image denoising methods usually assume that the noise exists in all bands, so they cannot effectively deal with this situation

[0005] However, the hyperspectral image denoising method based on the 2D image denoising method is extended to process each band completely independently, without considering the correlation between the bands, so the denoising effect is not ideal

The band-dependent hyperspectral image denoising method considers that the noise level of each band is equivalent, and does not consider the difference in the noise level of different bands, so it cannot achieve good results for the situation where the noise only appears in certain or specific bands

Images