A method for suppressing non-uniform background of wide-area star images using a CCD sensor

Abstract

This invention discloses a method for suppressing non-uniform background in wide-area star images using a CCD sensor. The method first describes the image format and uses this information to analyze the background imaging characteristics within the image. A fixed-window one-dimensional median filter is then used to estimate and suppress the overall non-uniform background of the star image. Finally, based on the concept of local contrast, an adaptive one-dimensional median filter is used on the processed star image to accurately estimate and suppress the background. This method effectively removes noise, non-uniformity, and bright lines caused by smear in the star image background, effectively suppressing the background while preserving the stars and targets on the star image.

Description

Technical Field

[0001] The present invention belongs to the technical field of computer vision, and in particular relates to a method for suppressing non-uniform background of a wide-area star image of a CCD sensor. Background Art

[0002] Background suppression technology is a method used to reduce or eliminate background interference in images. It is widely used in image processing and computer vision, particularly in target detection and tracking tasks. During the imaging of space targets, a large amount of noise appears in the image due to the influence of stray light in space and the detection equipment itself. This reduces image quality, affects the accuracy of target detection, and the precision of target centroid positioning. Therefore, image preprocessing is necessary to improve image quality and reduce the impact of noise on target detection and positioning. Classic denoising methods include image frequency domain filtering, mean filtering, median filtering, Gaussian filtering, Wiener filtering, and wavelet transform.

[0003] CCD image sensors can directly convert optical signals into digital electrical signals, feature low power consumption, stable performance, low noise, and high pixel integration. They are widely used in space target observation and reconnaissance, space debris warning, astronomical observation, and other fields. Research on wide-area star image background suppression primarily aims to support space target extraction. By processing the image before target detection, the aim is to suppress or eliminate non-uniform background noise. In star images, the size, shape, and grayscale values ​​of space targets are highly similar to those of noise. The image background is unevenly distributed due to stray light. For CCD detectors, bright lines caused by the smear effect appear in the image. These factors make traditional image denoising algorithms difficult to apply to star image denoising.

[0004] In recent years, based on the research results of star image background suppression at home and abroad, the processing of star image background suppression is roughly as follows:

[0005] 1) Data preprocessing: Preprocess the image, including denoising, smoothing, and grayscale operations, to improve the effect of subsequent processing;

[0006] 2) Background segmentation: Segment the image into target area and background area. Common techniques include threshold segmentation and edge detection.

[0007] 3) Background estimation: Identify and quantify the background signal in the star map by performing background estimation, usually using methods such as mean filtering and smoothing filtering;

[0008] 4) Background suppression: Subtract the established background model from the original star image data to suppress the background signal. This can be achieved through pixel-by-pixel subtraction or other mathematical calculations.

[0009] In the background suppression step of star images, the quality of background estimation determines the effectiveness of the background suppression. Through reasonable and accurate modeling and estimation, the system can more effectively identify foreground objects in the image, thereby achieving precise background suppression. Choosing the right background estimation method is crucial to the success of the entire background suppression process. A commonly used background estimation technique is statistically-based background estimation. This method typically plots a histogram of pixel brightness and then analyzes this distribution to determine statistical parameters of the background. Background estimation is achieved by setting a threshold. Because celestial objects and other bright spots are high-brightness outliers in the image, the statistical analysis focuses on dark pixels in the image—that is, only pixels below a certain brightness value are counted in the background estimation. The accuracy of statistical background estimation is significantly affected in star images with non-uniform illumination conditions. Furthermore, when removing the estimated background, the signals of faint targets may be incorrectly removed, especially those with brightness close to the background level. To address the problems of low background estimation accuracy and the easy removal of faint targets under non-uniform illumination conditions, this project proposes a non-uniform background suppression method for wide-area star images using CCD sensors. Background suppression is crucial for extracting faint targets in space. Summary of the Invention

[0010] To overcome the shortcomings of existing technologies, the present invention provides a method for suppressing non-uniform background in wide-area star images using a CCD sensor. The method first describes the image format and uses this information to analyze the background imaging characteristics within the image. A fixed-window one-dimensional median filter is then used to estimate and suppress the overall non-uniform background of the star image. Finally, an adaptive one-dimensional median filter, based on the principle of local contrast, is applied to the processed star image for precise background estimation and suppression. This method effectively removes noise, non-uniformity, and bright lines caused by smear in the star image background, effectively suppressing the background while preserving the stars and targets on the star image.

[0011] The technical solutions adopted by the present invention to solve the technical problems are as follows:

[0012] Step 1: Analyze the imaging characteristics of the starry sky observation image and describe the starry sky observation image as:

[0013] I(x,y)=I s (x,y)+I B (x,y)+I N (x,y)+I L (x,y)

[0014] Where I(x,y) is the observation image obtained by the CCD focal plane of the observation camera; (x,y) is the coordinate of the pixel point, and 1≤x≤M, 1≤y≤N, where M=4096, N=4096; I S (x, y) is the target image, including the observation image of stars and space targets; I B (x, y) is the starry sky background image; I N (x, y) is a noise image, including spatial radiation noise, CCD fixed noise, and dark background noise, and it is assumed that I N (x,y) obeys zero-mean Gaussian noise; I L (x,y) is the Smear image;

[0015] A one-dimensional median filter is performed on the original image using a window of size W0 to estimate the overall non-uniform background. The background estimate B0 is:

[0016]

[0017] By performing a difference with the original image, the image I0 obtained by removing the overall non-uniform noise in the background and the bright lines on the image caused by the Smear effect is:

[0018] I0(x,y)=I(x,y)-B0(x,y)

[0019] Where I0(x,y) is the grayscale value at the (x,y) coordinate position in image I0, and image I0 is the image obtained after the non-uniform background of the wide-area star map CCD sensor is initially suppressed;

[0020] Step 2: Count the grayscale value and T of each pixel in image I0, and calculate T according to the following formula:

[0021]

[0022] Where (x, y) is the coordinate of the pixel point, I0(x+i, y+j) is the grayscale value at the coordinate position (x+i, y+j) in image I0, and a is the local contrast value;

[0023] The local contrast T is calculated according to the following formula l :

[0024]

[0025] Local contrast T l Used to adaptively adjust the size of the variable window;

[0026] Step 3: Use the variable window size method to perform one-dimensional median filtering on the image I0 to estimate the non-uniform background; according to the local contrast T obtained in step 2 l, determine the size of the filter window W1, the threshold T t Determined by the following formula:

[0027]

[0028] Where μ B is the grayscale mean of the statistical image I0;

[0029] The non-uniform background estimation B1 of image I0 obtained by adaptive filtering is:

[0030]

[0031] Performing a difference with image I0 to remove the non-uniform background, the image I1 is obtained as follows:

[0032] I1(x,y)=I0(x,y)-B1(x,y)

[0033] Image I1 is the final image obtained after the non-uniform background of the wide-area star image of the CCD sensor is suppressed.

[0034] Furthermore, the one-dimensional median filter processing window W0 in step 1 is 128.

[0035] Furthermore, the local comparison range a in step 2 is 24.

[0036] Furthermore, in step 3, according to the local contrast T l With threshold T t The size of the filter window W1 is adjusted according to the size relationship. W1 represents the one-dimensional median filter window being processed with the center at (x, y). st The default initial window size is W1, W max The maximum window size allowed by W1; its adjustment steps are:

[0037] Step 3-1: Initial window W1 = W st , if the current one-dimensional median filter window centered at (x, y) is T l <T t , then the output window size is W st The filtering result is, otherwise go to step 3-2;

[0038] Step 3-2: If W1 <W max , then increase the window: W1 plus C0, go to step 3-1, otherwise go to step 3-3;

[0039] Step 3-3: At this time, W1=W max , then the output window size is W max The filtering result when .

[0040] Furthermore, the initial window size W st 24, W max is 48, and the constant C0 is 8.

[0041] The beneficial effects of the present invention are as follows:

[0042] The method of the present invention can effectively remove noise, non-uniformity and bright lines caused by smear in the background of the star map at the same time, and preserves the stars and targets on the star map while suppressing the background. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a diagram illustrating an implementation framework of a method for suppressing non-uniform background of a wide-area star image of a CCD sensor according to an embodiment of the present invention.

[0044] Figure 2 This is a schematic diagram of the one-dimensional median filtering principle of an embodiment of the present invention.

[0045] Figure 3 This is a wide-area starry sky observation image of the CCD sensor used in the embodiment of the present invention.

[0046] Figure 4 For the general Figure 3 The image is obtained after background suppression according to the method of the present invention. DETAILED DESCRIPTION

[0047] The present invention will be further described below with reference to the accompanying drawings and examples.

[0048] To overcome the shortcomings of existing star image background suppression algorithms, which often remove some stars and faint objects from the image while suppressing the background, this paper proposes a method for suppressing non-uniform background in wide-area star images from CCD sensors. This method effectively removes noise, non-uniformity, and bright lines caused by smear in the star image background. Compared to traditional algorithms, it estimates the non-uniform noise background more accurately and does not require prior image information.

[0049] A method for suppressing non-uniform background of a wide-area star image of a CCD sensor comprises the following steps:

[0050] ① Analyze the imaging characteristics of the starry sky observation image and describe the starry sky observation image as:

[0051] I(x,y)=I s (x,y)+I B (x,y)+I N (x,y)+I L (x,y)

[0052] Where I(x,y) is the observation image obtained by the CCD focal plane of the observation camera; (x,y) is the coordinate of the pixel point, and 1≤x≤M, 1≤y≤N, where M=4096, N=4096; I S (x, y) is the target image, including the observation image of stars and space targets; I B (x, y) is the starry sky background image; I N (x, y) is a noise image, including spatial radiation noise, CCD fixed noise, dark background noise, and is assumed to obey zero-mean Gaussian noise; I L (x,y) is the Smear image;

[0053] A one-dimensional median filter is performed on the original image using a window of size W0 to estimate the overall non-uniform background. The background estimate B0 is:

[0054]

[0055] By performing a difference with the original image, the image I0 is obtained by removing the overall non-uniform noise in the background and the bright lines on the image caused by the Smear effect.

[0056] I0(x,y)=I(x,y)-B0(x,y)

[0057] Image I0 is the image obtained after the initial suppression of the non-uniform background of the wide-area star map CCD sensor;

[0058] ② Count the grayscale value and T of each pixel of image I0, and calculate T according to the following formula:

[0059]

[0060] Where (x, y) is the coordinate of the pixel point, I0(x+i, y+j) is the grayscale value at the coordinate position (x+i, y+j) in image I0, and a is the local contrast value;

[0061] The local contrast T is calculated according to the following formula l :

[0062]

[0063] Where I0(x,y) is the grayscale value at the (x,y) coordinate position in image I0, and the local contrast T l Used to adaptively adjust the size of the variable window;

[0064] ③Use the variable window size method to perform one-dimensional median filtering on the image obtained in the previous step to accurately estimate the non-uniform background; according to the local contrast T obtained in ② l , determine the size of the filter window W1, the threshold Tt The threshold is determined by the following formula:

[0065]

[0066] Where I0(x,y) is the grayscale value at the (x,y) coordinate position, μ B is the grayscale mean of the statistical image I0;

[0067] The non-uniform background estimation B1 of image I0 obtained by adaptive filtering is:

[0068]

[0069] Difference is performed with image I0 to remove the non-uniform background, and the image I1 is obtained as

[0070] I1(x,y)=I0(x,y)-B1(x,y)

[0071] Image I1 is the final image obtained after the non-uniform background of the wide-area star image of the CCD sensor is suppressed.

[0072] According to the local contrast T l With threshold T t The size of the filter window W1 is adjusted according to the size relationship. W1 represents the one-dimensional median filter window being processed with the center at (x, y). st The default initial window size is W1, W max The maximum window size allowed by W1. The adjustment steps are:

[0073] a) Initial window W1 = W st , if the current one-dimensional median filter window centered at (x, y) is T l <T t , then the output window size is W st , otherwise go to b);

[0074] b) If W1 <W max , then increase the window: W1=W1+C0, go to a), otherwise go to c);

[0075] c) At this time, W1=W max , then the output window size is W max The filtering result when .

[0076] Initial window size W st 24, W max is 48, and the constant C0 is 8.

[0077] Example:

[0078] Figure 1This is a framework diagram of an implementation of a method for suppressing non-uniform background of a wide-area star image of a CCD sensor according to this embodiment. The specific implementation steps are as follows:

[0079] Step 1: Analyze the imaging characteristics of the input CCD sensor wide-area star map, process the image as a whole, obtain the overall non-uniform background estimation result of the image, and suppress it.

[0080] Generally speaking, during the imaging of space targets, a significant amount of noise is introduced due to the influence of stray light in space and the detection equipment itself. This reduces image quality, impacting target detection accuracy and tracking precision. Therefore, image preprocessing is necessary to improve image quality and reduce the impact of noise on target detection and tracking. The main sources of background noise include: space radiation noise, noise from the imaging system's detector itself, and dark background noise.

[0081] By analyzing the background imaging characteristics in the image, the star observation image is described as:

[0082] I(x,y)=I s (x,y)+I B (x,y)+I N (x,y)+I L (x,y)

[0083] Where I(x,y) is the observation image obtained by the CCD focal plane of the observation camera; (x,y) is the coordinate of the pixel point, and 1≤x≤M, 1≤y≤N, where M=4096, N=4096; I S (x, y) is the target image, including the observation image of stars and space targets; I B (x, y) is the starry sky background image; I N (x, y) is a noise image, including spatial radiation noise, CCD fixed noise, dark background noise, and is assumed to obey zero-mean Gaussian noise; I L (x,y) is the Smear image.

[0084] Based on the background imaging characteristics of the sky observation image, the original image columns contain targets and stars, resulting in higher grayscale values ​​in some pixels. At the same time, due to non-uniform noise contamination, the grayscale values ​​of each column of pixels in the image fluctuate significantly. Selecting a larger window can help estimate the non-uniform fluctuations of the background.

[0085] Therefore, one-dimensional median filtering is used for processing, and its principle is as follows Figure 2 As shown, at this time, the filter window size is W0=128, and the overall non-uniform background estimation is performed, then the background estimation B0 is;

[0086]

[0087] By performing a difference with the original image, the image I0 is obtained by removing the overall non-uniform noise in the background and the bright lines on the image caused by the Smear effect.

[0088] I0(x,y)=I(x,y)-B0(x,y)

[0089] Image I0 is the image obtained after the non-uniform background of the wide-area star map CCD sensor is initially suppressed.

[0090] Step 2: Count the grayscale value of each pixel of the image I0 obtained in step 1, and then calculate its local contrast to obtain the basis for the adaptive filtering window for estimating the local non-uniform background.

[0091] For star maps, most pixels in the image belong to the background and have low grayscale values. The grayscale value distribution range is relatively concentrated. At the same time, there are target and star pixels with higher grayscale values. Some pixels even have saturated grayscale values ​​and their distribution range is relatively dispersed.

[0092] Therefore, the grayscale value and T of each pixel of image I0 are counted, and T is calculated according to the following formula:

[0093]

[0094] Where (x, y) is the coordinate of the pixel point, I0(x+i, y+J) is the grayscale value at the coordinate position (x+i, y+j) in image I0, a is the local contrast range, a=24;

[0095] The local contrast T is calculated according to the following formula l :

[0096]

[0097] Where I0(x,y) is the grayscale value at the (x,y) coordinate position in image I0, and the local contrast T l Used to adaptively adjust the size of the variable window.

[0098] Step 3: Process the image accurately to obtain local non-uniform background estimation and suppress it.

[0099] Using a larger window can estimate the background non-uniformity fluctuations, but the overall background grayscale value curve is relatively flat and cannot accurately estimate the local non-uniformity. At the same time, median filtering requires sorting the grayscale values ​​of pixels in the window, and a large window will introduce a large amount of calculation. However, the smaller the window, the better. Although a small window can more accurately estimate the local non-uniformity in the background, if the window contains a target or star, if the target or star occupies more than 50% of the pixels in the small window, then the target or star pixels will be mistakenly estimated as background, resulting in the loss of target information during background subtraction.

[0100] Therefore, a variable window size method is used to perform one-dimensional median filtering on the image obtained in the previous step, and the residual local non-uniform noise in the image is further removed to accurately estimate the non-uniform background; according to the local contrast T obtained in step 2 l , determine the size of the filter window W1, the threshold T t The threshold is determined by the following formula:

[0101]

[0102] Where I0(x,y) is the grayscale value at the (x,y) coordinate position, μ B is the gray value of the statistical image I0;

[0103] According to the local contrast T l With threshold T t The size of the filter window W1 is adjusted according to the size relationship. W1 represents the one-dimensional median filter window being processed with the center at (x, y). st The default initial window size is W1, W st =24,W max is the maximum window size allowed by W1, W max =48, C0 is a constant, C0=8. The adjustment steps are:

[0104] a) Initial window W1 = W st , if the current one-dimensional median filter window centered at (x, y) is T l <T t , then the output window size is W st , otherwise go to b);

[0105] b) If W1 <W max , then increase the window: W1=W1+C0, go to a), otherwise go to c);

[0106] c) At this time, W1=W max , then the output window size is W max The filtering result when .

[0107] The non-uniform background estimation B1 of image I0 obtained by adaptive filtering is:

[0108]

[0109] Difference is performed with image I0 to remove the non-uniform background, and the image I1 is obtained as

[0110] I1(x,y)=I0(x,y)-B1(x,y)

[0111] Image I1 is the final image obtained after the non-uniform background of the wide-area star image of the CCD sensor is suppressed.

[0112] Figure 3 This is a wide-area starry sky observation image of a CCD sensor used in an embodiment of the present invention. Figure 4 For the general Figure 3 The image shown is the image after background suppression by the above method. It can be seen that the non-uniform background noise of the original image is well removed, and the low signal-to-noise ratio targets in the star image are retained.

Claims

1. A method for suppressing non-uniform background of wide-area star images of a CCD sensor, characterized in that: The steps include: Step 1: Analyze the imaging characteristics of the starry sky observation image and describe the starry sky observation image as: I(x,y)=I s (x,y)+I B (x,y)+I N (x,y)+I L (x,y) Where I(x,y) is the observation image obtained by the CCD focal plane of the observation camera; (x,y) is the coordinate of the pixel point, and 1≤x≤M, 1≤y≤N, where M=4096, N=4096; I S (x, y) is the target image, including the observation image of stars and space targets; I B (x, y) is the starry sky background image; I N (x, y) is a noise image, including spatial radiation noise, CCD fixed noise, and dark background noise, and it is assumed that I N (x,y) obeys zero-mean Gaussian noise; I L (x,y) is the Smear image; A one-dimensional median filter is performed on the original image using a window of size W0 to estimate the overall non-uniform background. The background estimate B0 is: By performing a difference with the original image, the image I0 obtained by removing the overall non-uniform noise in the background and the bright lines on the image caused by the Smear effect is: I0(x,y)=I(x,y)-B0(x,y) Where I0(x,y) is the grayscale value at the (x,y) coordinate position in image I0, and image I0 is the image obtained after the non-uniform background of the wide-area star map CCD sensor is initially suppressed; Step 2: Count the grayscale value and T of each pixel in image I0, and calculate T according to the following formula: Where (x, y) is the coordinate of the pixel point, I0(x+i, y+j) is the grayscale value at the coordinate position (x+i, y+j) in image I0, and a is the local contrast value; The local contrast T is calculated according to the following formula l : Local contrast T l Used to adaptively adjust the size of the variable window; Step 3: Use the variable window size method to perform one-dimensional median filtering on the image I0 to estimate the non-uniform background; according to the local contrast T obtained in step 2 l , determine the size of the filter window W1, the threshold T t Determined by the following formula: Where μ B is the grayscale mean of the statistical image I0; The non-uniform background estimation B1 of image I0 obtained by adaptive filtering is: Performing a difference with image I0 to remove the non-uniform background, the image I1 is obtained as follows: I1(x,y)=I0(x,y)-B1(x,y) Image I1 is the final image obtained after the non-uniform background of the wide-area star image of the CCD sensor is suppressed.

2. The method for suppressing non-uniform background of a wide-area star image of a CCD sensor according to claim 1, characterized in that: The one-dimensional median filter processing window W0 in step 1 is 128.

3. The method for suppressing non-uniform background of a wide-area star image of a CCD sensor according to claim 1, characterized in that: The local comparison range a in step 2 is 24.

4. The method for suppressing non-uniform background of a wide-area star image of a CCD sensor according to claim 1, characterized in that: In step 3, the local contrast T l With threshold T t The size of the filter window W1 is adjusted according to the size relationship. W1 represents the one-dimensional median filter window being processed with the center at (x, y). st The default initial window size is W1, W max The maximum window size allowed for W1; The adjustment steps are: Step 3-1: Initial window W1 = W st , if the current one-dimensional median filter window centered at (x, y) is T l <T t , then the output window size is W st The filtering result is, otherwise go to step 3-2; Step 3-2: If W1 <W max , then increase the window: W1 plus C0, go to step 3-1, otherwise go to step 3-3; Step 3-3: At this time, W1=W max , then the output window size is W max The filtering result when .

5. The method for suppressing non-uniform background of a wide-area star image of a CCD sensor according to claim 4, characterized in that: The initial window size W st 24, W max is 48, and the constant C0 is 8.

Images