A hyperspectral image denoising method, system and medium based on adaptive rank correction

Abstract

The invention discloses a hyperspectral image denoising method, system and medium based on adaptive rank correction. The hyperspectral image denoising method of the present invention aims at the cleanness of each matrixed block degraded hyperspectral image and the initial estimation of the matrixed block. A hyperspectral image denoising model based on adaptive rank correction is established for hyperspectral images, and the adaptive rank correction penalty function is used to adaptively offset the penalty of the nuclear norm for large singular values. Based on l 2,1 The norm eliminates sparse noise, and the alternating direction multiplier algorithm is used to solve the block clean image and combine to obtain the final estimated clean hyperspectral image. The present invention can solve the problems that most of the existing hyperspectral denoising methods based on low-rank matrix recovery use nuclear norm to approximate the matrix rank, which leads to excessive punishment of large singular values ​​and fixed rank information, which needs to be defined in advance, making the method inflexible, etc., and can eliminate Complex noises such as Gaussian noise, impulse noise, dead lines and banding in hyperspectral images.

Description

technical field [0001] The invention relates to hyperspectral image denoising technology, in particular to a hyperspectral image denoising method, system and medium based on adaptive rank correction. Background technique [0002] Hyperspectral imagery (HSI) contains rich spectral information and is widely used in environmental monitoring, urban planning, geological exploration, and agricultural and forestry production. During the imaging and transmission process, due to the influence of many complex factors, the acquired hyperspectral image contains various types of mixed noise, such as Gaussian noise, impulse noise, banding, dead line, etc. These complex mixed noises not only affect image quality, but also cause great trouble to subsequent applications. Therefore, proposing an effective hyperspectral image denoising method is of great significance for hyperspectral image processing analysis and subsequent applications. [0003] Traditional 2D image denoising methods, such...

AI Technical Summary

Problems solved by technology

[0006] The technical problem to be solved by the present invention: In order to solve most of the existing hyperspectral denoising methods based on low-rank matrix restoration, the nuclear norm is used to approximate the matrix rank, which leads to excessive punishment. The large singular value and rank r information are fixed and need to be defined in advance so that the method Inflexible and other problems, provide a hyperspectral image denoising method, system and medium based on Adaptive Rank Correction (ARC), the invention can remove Gaussian noise, impulse noise, dead lines and bands and other complex noise

Images