Hyperspectral image denoising method based on multi-target low-rank sparsity and space spectrum total variation

Abstract

The invention relates to a hyperspectral image denoising method based on multi-target low-rank sparsity and space spectrum total variation. In combination with a multi-objective optimization theory, ahyperspectral image denoising recovery problem is converted into a multi-objective optimization problem of a sparse noise image item, a low-rank clean image item and a data fidelity item in combination with a spatial-spectral total variation item. The sparse noise is modeled by using the 10 norm, and the low-rank term is modeled by using the nuclear norm. Then, the three target items are optimized at the same time through the strong optimizing capacity of the multi-target evolutionary algorithm, and the model is solved to achieve the optimal set of solutions. According to the method, the problems of inaccurate l1 norm sparsity modeling and difficulty in optimal regularization parameter selection in an existing hyperspectral image sparse denoising method can be solved, and in addition, a sub-fitness updating strategy is designed, so that the algorithm is carried out more effectively. According to the invention, the applicability and precision of hyperspectral image denoising can be effectively improved.

Description

technical field [0001] The invention is based on the field of remote sensing image technology processing, and in particular relates to a hyperspectral image denoising method based on multi-target low-rank sparse and spatial spectrum full variation. Background technique [0002] Hyperspectral imagery plays a key role in many applications due to its rich spectral features, such as fine classification of crops and mineral identification. However, due to the hardware influence of the imaging sensor or the interference of external factors, it is difficult to avoid the existence of noise on the hyperspectral image, such as typical Gaussian noise and band noise. These noises greatly affect the subsequent interpretation and application of images, such as image classification and change detection, which are affected by noise. Therefore, how to restore clean hyperspectral remote sensing images has become one of the research hotspots in the field of remote sensing image processing. ...

AI Technical Summary

Problems solved by technology

Therefore, this method faces two main problems: (1) using l 1 Inaccurate modeling of norm-sparse noise

However, due to the complexity of remote sensing images, this condition is often difficult to fully satisfy, resulting in inaccurate sparse noise modeling; (2) The optimal regularization parameter is difficult to be determined automatically by humans, and it takes a lot of manpower to select this weight parameter

Due to the existence of multiple optimization items such as low-rank items, sparse items, spatial variation items, and data fidelity items in the sparse low-rank denoising model, weight parameters are usually introduced in existing methods to combine them into an objective function, where The selection of sensitive weight parameters greatly affects the accuracy of hyperspectral image denoising

Images