CN-121767227-B - Near-field electromagnetic wave imaging multi-mode noise cooperative suppression method

CN121767227BCN 121767227 BCN121767227 BCN 121767227BCN-121767227-B

Abstract

The invention discloses a near-field electromagnetic wave imaging multi-mode noise cooperative suppression method, relates to the technical field of electromagnetic wave imaging, and aims to solve the problems that speckle, stripe and Gaussian mixture noise suppression effects are poor and a target structure is easy to lose in the prior art. The method adopts a three-stage collaborative scheme of complex domain speckle suppression, multidirectional stripe separation and cross-channel Gaussian suppression, firstly, the amplitude and the phase of a complex domain image are processed through a self-adaptive threshold value to remove speckle noise, then horizontal/vertical ADOM filtering is respectively carried out on the real part and the imaginary part of the complex domain to remove stripe noise, finally, the BM3D filtering and three-dimensional transformation domain are combined to optimize and suppress Gaussian noise, and robustness is improved through multi-frequency point fusion. Experiments show that the SSIM is improved by 21.7% compared with the traditional method, the GSSIM and the PSNR are optimal, the target structural characteristics can be reserved in a strong noise environment, and the method is suitable for scenes such as defect detection of near-field electromagnetic wave synthetic aperture imaging.

Inventors

ZHAO CHUNYING
LIANG JING
DU YUANJIE
WANG FENG
PENG LIBIAO
ZHAO XINGZHONG
TANG YU
BI DONGJIE
LI XIFENG
XIE YONGLE
ZENG SHICHAO
ZHANG QINLEI

Assignees

成都天奥技术发展有限公司
电子科技大学

Dates

Publication Date: 20260512
Application Date: 20260303

Claims (5)

1. The near-field electromagnetic wave imaging multi-mode noise cooperative suppression method is characterized by comprising the following steps of: s1, complex domain speckle noise suppression is carried out, and specifically the following steps are carried out: s101, acquiring an original near-field electromagnetic wave imaging image containing noise Wherein The representation of the complex number field is provided, Separating the original near-field electromagnetic wave imaging image for the image pixel dimension Amplitude component of (a) And phase component ; S102, based on the amplitude component Global mean of (2) With global standard deviation The amplitude adaptive threshold is constructed as follows: ; Based on the phase component Global mean of (2) With global standard deviation The phase adaptive threshold is constructed as follows: ; s103, according to the self-adaptive threshold value, the amplitude component is subjected to Phase component Respectively performing nonlinear inhibition processing to obtain amplitude after speckle noise inhibition And phase of ; S104, based on And (3) with Reconstructing complex domain images Wherein Is an imaginary unit; s2, complex domain stripe noise suppression is carried out, and specifically the following steps are carried out: s201, pair The real part of (2) Normalized to Performing filtering in horizontal direction ADOM to obtain Filtering in vertical direction ADOM to obtain The ADOM filters satisfy: ; ; Wherein the method comprises the steps of As the direction differential operator, 、 Respectively the gradients in the horizontal and vertical directions, Is a regularization parameter; s202, pair Imaginary part of (2) Normalized to [0,255], and the filtering result of the imaginary part in the horizontal direction is obtained according to the ADOM filtering mode And the imaginary part filtering result in the vertical direction ; S203, reconstructing the complex domain image after the stripe noise suppression ; S3, cross-channel Gaussian noise suppression is carried out, and specifically the method comprises the following steps: S301, pair The real part of (2) Execution of Filtering to obtain For a pair of Imaginary part of (2) Performing BM3D filtering Wherein Representing a block matching 3D filtering operation; s302, pair And (3) with Performs three-dimensional wavelet packet transforms on a union of (a) and (b) a transform By adaptive threshold operators After processing, the three-dimensional wavelet packet is transformed by inverse Obtaining a final output image I.e. ; S4, multi-frequency point fusion optimization, which comprises the following steps: and (3) respectively executing phase alignment of coherent fusion, amplitude superposition and pixel-level weighted average of incoherent fusion on the images processed in the steps (S1) to (S3) under different frequency points, so as to realize comprehensive suppression of multi-modal noise.
2. The near field electromagnetic wave imaging multi-mode noise cooperative suppression method according to claim 1, wherein the specific formula of the nonlinear suppression processing in step S1 is as follows: ; ; Wherein the method comprises the steps of Is the image pixel coordinates.
3. The near field electromagnetic wave imaging multi-mode noise cooperative suppression method according to claim 1, wherein a specific formula of ADOM filtering of an imaginary part in the step S2 is as follows: ; ; Wherein the method comprises the steps of 、 Representing the imaginary part of the image in the horizontal direction and in the vertical direction respectively, As the direction differential operator, And Respectively represent the gradient in the horizontal direction and the gradient in the vertical direction, Is a regularization parameter.
4. The near field electromagnetic wave imaging multi-modal noise cooperative suppression method as claimed in claim 1, wherein the BM3D filtering in step S3 is performed by performing block matching grouping on real part/imaginary part, performing 3D conversion and thresholding on each group of blocks, and performing inverse 3D conversion and block recombination to obtain And (3) with The adaptive threshold operator Threshold of (2) According to the self-adaptive adjustment of the noise intensity of the image, the higher the noise intensity is The larger the value is.
5. The near-field electromagnetic wave imaging multi-modal noise cooperative suppression method as claimed in claim 1, wherein the multi-frequency point fusion optimization in step S4 is specifically implemented by adopting a least square phase alignment algorithm for high-frequency band images during coherent fusion, adopting a weighting coefficient based on image definition during incoherent fusion, wherein the higher the definition is, the larger the pixel weight is, and the weighting coefficient range is 。

Description

Near-field electromagnetic wave imaging multi-mode noise cooperative suppression method Technical Field The invention relates to the technical field of electromagnetic wave imaging, in particular to a near-field electromagnetic wave imaging multi-mode noise cooperative suppression method. Background Electromagnetic wave imaging can penetrate through various media due to unique physical characteristics, and provides a high-resolution imaging effect, however, in the practical application process, an electromagnetic wave imaging system is often affected by various noises, which not only seriously interfere with the definition of an imaging image, but also affect the subsequent tasks of image processing, target detection, identification and the like, particularly in near-field electromagnetic wave imaging, the noise types caused by system and environmental factors are complex, the precision of an imaging result is limited due to the existence of the noise, and therefore, how to effectively remove the noise on the premise of keeping the structural characteristics of the image is an important task in near-field electromagnetic wave imaging technical research. The traditional near-field electromagnetic wave imaging denoising technology mainly adopts traditional filtering methods (such as mean filtering, median filtering and wavelet threshold denoising) and end-to-end image enhancement models (such as convolutional neural networks and generation countermeasure networks) based on deep learning, the methods can suppress Gaussian noise, pretzel noise and system thermal noise to a certain extent, but have the common limitations that the traditional methods are easy to excessively smooth details, so that edge and weak target characteristics are lost, the deep learning model depends on a large amount of labeling data, the generalization capability in an actual scene is limited, the non-uniform background interference, multipath effect and phase noise suppression effect on complex environments are poor, in addition, most algorithms do not fully consider the physical characteristics of electromagnetic wave imaging, the balance of structure fidelity and signal-to-noise ratio improvement is difficult to be maintained under strong noise, and the imaging precision and the subsequent automatic recognition performance are restricted. Disclosure of Invention Aiming at the defects of the prior art, the invention provides a near-field electromagnetic wave imaging multi-mode noise cooperative suppression method, which solves the problems in the background art. In order to achieve the purpose, the invention is realized by the following technical scheme that the near-field electromagnetic wave imaging multi-mode noise cooperative suppression method comprises the following steps: s1, complex domain speckle noise suppression is carried out, and specifically the following steps are carried out: s101, acquiring an original near-field electromagnetic wave imaging image containing noise WhereinThe representation of the complex number field is provided,Separating the original near-field electromagnetic wave imaging image for the image pixel dimensionAmplitude component of (a)And phase component; S102, based on the amplitude componentGlobal mean of (2)With global standard deviationThe amplitude adaptive threshold is constructed as follows: Based on the phase component Global mean of (2)With global standard deviationThe phase adaptive threshold is constructed as follows: s103, according to the self-adaptive threshold value, the amplitude component is subjected to Phase componentRespectively performing nonlinear inhibition processing to obtain amplitude after speckle noise inhibitionAnd phase of; S104, based onAnd (3) withReconstructing complex domain imagesWhereinIs an imaginary unit; s2, complex domain stripe noise suppression is carried out, and specifically the following steps are carried out: s201, pair The real part of (2)Normalized toPerforming filtering in horizontal direction ADOM to obtainFiltering in vertical direction ADOM to obtainThe ADOM filters satisfy: Wherein the method comprises the steps of As the direction differential operator,、Respectively the gradients in the horizontal and vertical directions,Is a regularization parameter; s202, pair Imaginary part of (2)Normalized to [0,255], and the filtering result of the imaginary part in the horizontal direction is obtained according to the ADOM filtering modeAnd the imaginary part filtering result in the vertical direction; S203, reconstructing the complex domain image after the stripe noise suppression; S3, cross-channel Gaussian noise suppression is carried out, and specifically the method comprises the following steps: S301, pair The real part of (2)Execution ofFiltering to obtainFor a pair ofImaginary part of (2)Performing BM3D filteringWhereinRepresenting a block matching 3D filtering operation; s302, pair And (3) withPerforms three-dimensional wavelet packet transforms on a