Noise suppression methods for long exposure images of detectors

By combining an improved maximum value filtering algorithm with the total variational method, dark current and non-uniform background noise in long-exposure images of the detector are filtered out, solving the problem of unsatisfactory noise processing in existing technologies and achieving better noise reduction results.

CN117391992BActive Publication Date: 2026-11-10CHANGCHUN INST OF OPTICS FINE MECHANICS & PHYSICS CHINESE ACAD OF SCI
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202311286217.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-10-07
Publication Date
2026-11-10
Estimated Expiration
2043-10-07

AI Technical Summary

Technical Problem

Existing long-exposure image noise suppression methods are sensitive to complex background noise and have unsatisfactory denoising effects, especially dark current noise and non-uniform background noise.

Method used

An improved maximum value filtering algorithm is used to filter out dark current noise, combined with the total variation method to suppress non-uniform background noise. Noise-contaminated pixels are filtered out by calculating local contrast and threshold, and the total variation algorithm is used to smooth the image background.

Benefits of technology

It accurately estimates and replaces pixels affected by dark current noise, effectively suppresses non-uniform background noise, and significantly improves image quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117391992B_ABST
    Figure CN117391992B_ABST
Patent Text Reader

Abstract

This invention relates to the field of image processing technology, and more particularly to a noise suppression method for long-exposure images of detectors. The method suppresses dark current noise and non-uniform background noise in long-exposure images of detectors, specifically including the following steps: S1, filtering out dark current noise: using an improved maximum value filtering algorithm to filter out pixels contaminated by dark current noise in the original image, obtaining a preprocessed image; S2, suppressing non-uniform background noise: using a total variation algorithm to filter out non-uniform background noise in the preprocessed image. This invention can accurately estimate and replace pixels affected by dark current noise, and uses a total variation method to suppress non-uniform background noise, resulting in good denoising performance.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method for noise suppression of long-exposure images of detectors. Background Technology

[0002] With the continuous development of aerospace technology, the strategic importance of space information in political, military, and economic fields is increasing. Currently, the main method for acquiring space information is to use space cameras to monitor space targets and obtain important parameters of spacecraft that pose a threat to aerospace systems, including the spacecraft's mission, size, shape, and orbital parameters. To detect more subtle space information, the exposure time of the detector is usually increased to accumulate more light energy from faint space targets. However, when the exposure time is increased to the second level or higher, the dark current noise and non-uniform background noise of the detector will seriously affect the image quality.

[0003] Existing methods for suppressing non-uniform background noise are mainly divided into two categories: calibration-based methods and scene-based methods. Calibration-based methods are computationally simple and have high correction accuracy; however, as detector operating time increases, temperature and time drift, as well as complex spatial target backgrounds, render the correction parameters inapplicable. Scene-based methods estimate noise based on the difference between the target and the background, without considering temporal correlation. There are two main processing methods: frequency domain filtering and spatial domain filtering. However, while frequency domain filtering can eliminate noise, it also introduces some distortion to the target. For images with star-map backgrounds, target edges are prone to oscillation after frequency domain filtering. Spatial domain filtering is a neighborhood processing method that directly processes pixels within the image space to smooth or sharpen the image. However, both methods are sensitive to complex background noise, and their denoising effects are not ideal. Summary of the Invention

[0004] To address the problem that existing noise suppression methods for long-exposure images are sensitive to complex background noise and have unsatisfactory denoising effects, this invention provides a noise suppression method for detector long-exposure images. This method can accurately estimate and replace pixels affected by dark current noise, and uses the total variation method to suppress non-uniform background noise, resulting in better denoising performance.

[0005] The noise suppression method for long-exposure images of detectors provided by this invention is used to suppress dark current noise and non-uniform background noise in long-exposure images of detectors, and specifically includes the following steps:

[0006] S1. Filter out dark current noise: The improved maximum value filtering algorithm is used to filter out pixels in the original image that are contaminated by dark current noise to obtain a preprocessed image.

[0007] Step S1 specifically includes the following steps:

[0008] S11. Calculate the mean gray value of the 8-neighborhood pixels at all pixels in the original image using the following formula:

[0009]

[0010] Where M(x,y) is the average gray value of the 8 neighboring pixels at pixel (x,y), and h i Let be the gray value of the 8-neighbor pixels at pixel (x,y), m be the pixel in the m-th row of the original image, and n be the pixel in the n-th column of the original image;

[0011] S12, Combining equation (1), calculate the local contrast of all pixels in the original image one by one using the following formula:

[0012]

[0013] Where c(x,y) is the local contrast of pixel (x,y), and I(x,y) is the gray value of pixel (x,y).

[0014] S13. Using an improved 8-neighborhood-based adaptive method, obtain the threshold corresponding to each pixel:

[0015]

[0016]

[0017] Th C (x,y)=μ C (x,y)+ασ C (x,y)x∈[1,M],y∈[1,N] (5);

[0018] Where, μ C (x,y) is the mean of the local contrast of the 8 neighboring pixels at pixel (x,y), σ C (x,y) represents the standard deviation of the local contrast of the 8 neighboring pixels at pixel (x,y), α is a coefficient, α=1.5, N=8;

[0019] S14. Pixels with local contrast greater than a threshold are identified as pixels contaminated by dark current noise. An improved maximum value filtering algorithm is used to filter out all pixels contaminated by dark current noise, resulting in the preprocessed image SLCM(x,y).

[0020]

[0021] Where W is a 3×3 maximum value filter template;

[0022] S2. Suppress non-uniform background noise: Use the total variation algorithm to filter out non-uniform background noise in the preprocessed image.

[0023] Preferably, step S2 specifically includes the following steps:

[0024] S21. Establish the constraint equation E[f(x,y)] using the total variational method:

[0025]

[0026]

[0027] Where f(x,y) is the gray value of the uncontaminated image at pixel (x,y). To contaminate the grayscale value of the image at pixel (x,y), For total variation, Here, λ is the regularization term, and λ is the regularization coefficient. It represents the absolute difference between the (x,y) pixel and its neighboring pixels;

[0028] S22. The gradient projection algorithm is used to solve the constraint equations and smooth the background of the preprocessed image.

[0029] Preferably, step S22 specifically includes the following steps:

[0030] S221. Define the polynomial in the integral term of equation (7) as a functional F:

[0031]

[0032] S222. By using the Euler-Lagrange equations to derive equation (9), we can obtain:

[0033]

[0034] Among them, f n+1 f represents the (n+1)th grayscale value of the preprocessed image. n The nth grayscale value of the preprocessed image. For the preprocessed image, Δt is a time auxiliary parameter, Δt = 1, and λ is 0.03;

[0035] S223. Iterate equation (10) 100 times to obtain the iteration result, which is the calculation result of the constraint equation.

[0036] Compared with the prior art, the present invention can achieve the following beneficial effects:

[0037] This invention can accurately estimate and replace pixels affected by dark current noise through an improved maximum value filtering algorithm. At the same time, this invention uses the total variation method to suppress non-uniform background noise such as residual thermal noise and detector readout noise in the image, smoothing the image background and achieving good noise reduction effect. Attached Figure Description

[0038] Figure 1 (a) is the original image of the detector provided according to an embodiment of the present invention in long exposure mode;

[0039] Figure 1 (b) is a magnified partial image of the star-shaped target in the original image provided according to an embodiment of the present invention;

[0040] Figure 1 (c) is a magnified image of a moving target in the original image provided according to an embodiment of the present invention;

[0041] Figure 1 (d) is a magnified image of dark current noise in the original image provided according to an embodiment of the present invention;

[0042] Figure 2 This is a flowchart illustrating a noise suppression method for long-exposure images of a detector according to an embodiment of the present invention.

[0043] Figure 3 This is a flowchart of an algorithm for noise suppression of long-exposure images of a detector according to an embodiment of the present invention. Detailed Implementation

[0044] In the following description, embodiments of the invention will be described with reference to the accompanying drawings. In the description below, the same modules are denoted by the same reference numerals. Where the same reference numerals are used, their names and functions are also the same. Therefore, their detailed description will not be repeated.

[0045] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not constitute a limitation thereof.

[0046] Figure 1 Image (a) shows the original image of the detector provided according to an embodiment of the present invention in long exposure mode; Figure 1 (b) shows a magnified partial image of the star-shaped target in the original image provided according to an embodiment of the present invention; Figure 1 (c) shows a magnified partial image of a moving target in the original image provided according to an embodiment of the present invention; Figure 1Image (d) shows a magnified local image of dark current noise in the original image provided according to an embodiment of the present invention.

[0047] like Figure 1 As shown in (a)-(d), the detector introduces dark current and non-uniform background in the 3s long exposure mode. Both star targets and moving targets are affected by dark current noise and non-uniform background noise, which affects our judgment of space targets. Therefore, noise suppression of long exposure images of the detector is crucial.

[0048] The noise suppression method for long exposure images of detectors proposed in this invention consists of two steps: the first step is to filter out dark current noise in the original image using an improved maximum value filtering algorithm to obtain a preprocessed image; the second step is to filter out non-uniform background noise in the preprocessed image using a total variation algorithm.

[0049] Figure 2 The flowchart of a noise suppression method for long-exposure images of a detector provided according to an embodiment of the present invention is shown; Figure 3 The algorithm flow for noise suppression of long-exposure images of a detector provided according to an embodiment of the present invention is shown.

[0050] like Figures 2-3 As shown, the noise suppression method for long-exposure images of detectors provided in this embodiment of the invention specifically includes the following steps:

[0051] S1. Filter out dark current noise: The improved maximum value filtering algorithm is used to filter out pixels in the original image that are contaminated by dark current noise to obtain a preprocessed image.

[0052] Step S1 specifically includes the following steps:

[0053] S11. Calculate the mean gray value of the 8-neighborhood pixels at all pixels in the original image using the following formula:

[0054]

[0055] Where M(x,y) is the average gray value of the 8 neighboring pixels at pixel (x,y), and h i Let be the gray value of the 8-neighbor pixels at pixel (x,y), m be the pixel in the m-th row of the original image, and n be the pixel in the n-th column of the original image;

[0056] S12, Combining equation (1), calculate the local contrast of all pixels in the original image one by one using the following formula:

[0057]

[0058] Where c(x,y) is the local contrast of pixel (x,y), and I(x,y) is the gray value of pixel (x,y).

[0059] S13. Using an improved 8-neighborhood-based adaptive method, obtain the threshold corresponding to each pixel:

[0060]

[0061]

[0062] Th C (x,y)=μ C (x,y)+ασ C (x,y)x∈[1,M],y∈[1,N] (5);

[0063] Where, μ C (x,y) is the mean of the local contrast of the 8 neighboring pixels at pixel (x,y), σ C (x,y) represents the standard deviation of the local contrast of the 8 neighboring pixels at pixel (x,y), α is a coefficient, α=1.5, N=8;

[0064] S14. Pixels with local contrast greater than a threshold are identified as pixels contaminated by dark current noise. An improved maximum value filtering algorithm is used to filter out all pixels contaminated by dark current noise, resulting in the preprocessed image SLCM(x,y).

[0065]

[0066] Where W is a 3×3 maximum value filter template;

[0067] S2. Suppress non-uniform background noise: Use the total variation algorithm to filter out non-uniform background noise in the preprocessed image.

[0068] Step S2 specifically includes the following steps:

[0069] S21. Establish the constraint equation E[f(x,y)] using the total variational method:

[0070]

[0071]

[0072] Where f(x,y) is the gray value of the uncontaminated image at pixel (x,y). To contaminate the grayscale value of the image at pixel (x,y), For total variation, Here, λ is the regularization term, and λ is the regularization coefficient. It represents the absolute difference between the (x,y) pixel and its neighboring pixels;

[0073] S22. The gradient projection algorithm is used to solve the constraint equations and smooth the background of the preprocessed image.

[0074] Step S22 specifically includes the following steps:

[0075] S221. Define the polynomial in the integral term of equation (7) as a functional F:

[0076]

[0077] S222. By using the Euler-Lagrange equations to derive equation (9), we can obtain:

[0078]

[0079] Among them, f n+1 f is the grayscale value of the (n+1)th image. n Let n be the grayscale value of the image. For image preprocessing, Δt is a time auxiliary parameter used to transform the problem of solving static nonlinear partial differential equations into an iterative process of dynamic partial differential equations. Δt = 1, and λ is 0.03.

[0080] The derivation process of step S222 is as follows: Given And the Euler-Lagrange equations are:

[0081]

[0082] Substituting equation (9) into equation (10), we get:

[0083]

[0084] in,

[0085]

[0086]

[0087]

[0088]

[0089]

[0090] Discretizing equations (12-a) to (12-e) in two-dimensional space yields:

[0091]

[0092]

[0093] f xx =f i+1,j +f i-1,j -2f i,j (13-c);

[0094] f yy =f i,j+1 +f i,j-1 -2f i,j (13-d);

[0095]

[0096] Equation (10) can be obtained by numerical calculation.

[0097] S223. Iterate equation (10) 100 times to obtain the iteration result, which is the calculation result of the constraint equation.

[0098] The number of iterations in equation (10) can be set according to the needs of image processing. The more iterations, the smaller the value of the constraint equation.

[0099] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this invention can be achieved, and this is not limited herein.

[0100] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A noise suppression method for long-exposure images of a detector, used to suppress dark current noise and non-uniform background noise in long-exposure images of a detector, characterized in that, Specifically, the steps include the following: S1. Filter out dark current noise: The improved maximum value filtering algorithm is used to filter out pixels in the original image that are contaminated by dark current noise to obtain a preprocessed image. Step S1 specifically includes the following steps: S11. Calculate the average gray value of the 8-neighborhood pixels at all pixels in the original image using the following formula: Where M(x,y) is the average gray value of the 8 neighboring pixels at pixel (x,y), and h i Let m be the gray value of the 8-neighbor pixels at pixel (x,y), m be the pixel in the m-th row of the original image, and n be the pixel in the n-th column of the original image. S12, Combining equation (1), calculate the local contrast of all pixels in the original image one by one using the following equation: Where c(x,y) is the local contrast of pixel (x,y), and I(x,y) is the gray value of pixel (x,y). S13. Using an improved 8-neighborhood-based adaptive method, obtain the threshold corresponding to each pixel: Th C (x,y)=μ C (x,y)+ασ C (x,y)x∈[1,M],y∈[1,N] (5); Where, μ C (x,y) is the mean of the local contrast of the 8 neighboring pixels at pixel (x,y), σ C (x,y) represents the standard deviation of the local contrast of the 8 neighboring pixels at pixel (x,y), α is a coefficient, α=1.5, N=8; S14. Single pixels with local contrast greater than a threshold are considered as pixels contaminated by dark current noise. All pixels contaminated by dark current noise are filtered out using an improved maximum value filtering algorithm to obtain the preprocessed image SLCM(x,y): Where W is a 3×3 maximum value filter template; S2. Suppress non-uniform background noise: Use the total variation algorithm to filter out non-uniform background noise in the preprocessed image.

2. The noise suppression method for long-exposure images of a detector according to claim 1, characterized in that, Step S2 specifically includes the following steps: S21. Establish the constraint equation E[f(x,y)] using the total variational method: Where f(x,y) is the gray value of the uncontaminated image at pixel (x,y). To contaminate the grayscale value of the image at pixel (x,y), For total variation, Here, λ is the regularization term, and λ is the regularization coefficient. It represents the absolute difference between the (x,y) pixel and its neighboring pixels; S22. The constraint equations are solved using a gradient projection algorithm to smooth the background of the preprocessed image.

3. The noise suppression method for long-exposure images of a detector according to claim 2, characterized in that, Step S22 specifically includes the following steps: S221. Define the polynomial in the integral term of equation (7) as a functional F: S222. By using the Euler-Lagrange equations to derive equation (9), we can obtain: Among them, f n+1 f represents the (n+1)th grayscale value of the preprocessed image. n The nth grayscale value of the preprocessed image. For the preprocessed image, Δt is a time auxiliary parameter, Δt = 1, and λ is 0.03; S223. Iterate equation (10) 100 times to obtain the iteration result, which is the calculation result of the constraint equation.