Hyperspectral remote sensing image strip removing method based on low-rank tensor multi-mode approximation

CN119228681AActive Publication Date: 2024-12-31BEIJING UNIV OF TECH
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202411769939.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-12-04
Publication Date
2024-12-31
Estimated Expiration
2044-12-04

AI Technical Summary

Technical Problem

The existing hyperspectral remote sensing image de-banding technology has problems such as noise residue and image oversmoothness, especially when dealing with densely distributed random banding, and the supervised learning method relies on the truth map label and has poor results, based on variation The smooth regularization method of fractions is difficult to avoid the step effect, and the low-rank tensor regularization method cannot effectively characterize the null spectrum consistency of the stereo band.

Method used

A hyperspectral remote sensing image de-banding method based on multimodal approximation of low-rank tensors is proposed. By expanding the weighted kernel norm and the truncated kernel norm to the tensor domain, combined with the alternating iterative multiplier method, the model is optimized to achieve clean images Accurate separation from band noise, maintaining the null spectral consistency of hyperspectral data and the characteristics of parameter adjustment.

Benefits of technology

Under the premise of maintaining the null spectral consistency of hyperspectral images, the parameter-free parameter-regulating characterization and highly linear characterization of banding noise are realized for spectral redundancy, effectively eliminating banding noise in hyperspectral images and providing a parameter-regulating character-regulating characterization The de-banding scheme improves the null spectrum consistency and noise separation performance of the image.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119228681A_ABST
    Figure CN119228681A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of hyperspectral remote sensing image denoising, in particular to a low-rank tensor multi-mode approximation-based hyperspectral remote sensing image strip removal method, which comprises the following steps of: dividing each pixel in observation hyperspectral data # imgabs0 # by the maximum value of all pixels to realize the amplitude normalization of the observation hyperspectral data; providing a weighted nuclear norm WNN-TnSVD based on a tensor modulus-n SVD, wherein the weighted nuclear norm WNN-TnSVD is used for constraining the spectrum redundancy of the potential hyperspectral image; a tensor SVD-based truncation nuclear norm TNN-TSVD is provided and is used for limiting the rank 1 property of the three-dimensional stripe component; establishing a hyperspectral image strip removal model based on low-rank tensor multi-modal approximation; on the basis of an alternating direction multiplier method, after the strip removal model is split into several simple sub-problems, a clean hyperspectral image and strip noise are obtained by alternately solving the sub-problems, and image and strip separation is achieved. The method is used for eliminating stripe noise and generating a clean hyperspectral image, and provides effective data for a watershed hyperspectral remote sensing observation task.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Hyperspectral image denoising method and system based on adaptive rank correction, and medium

    CN110458777A

  • Hyperspectral image denoising method based on non-local similarity and weighted truncation nuclear norm

    CN111951181A

  • Mixed noise hyperspectral image restoration method combining double low ranks and spatial spectrum total variation

    CN115439344A

  • Weighted adaptive tensor data complementation method based on DCT (Discrete Cosine Transform)

    CN116912107A

  • Hyperspectral image noise reduction method based on low-rank tensor two-factor joint optimization

    CN117876250A