A sunspot fine structure enhancement method based on linear Gaussian filtering

Abstract

The present application relates to a kind of solar filament fine structure enhancement method based on line Gauss filter, belong to image enhancement field.To overcome the current H-alpha image in solar filament fiber because of low contrast, leading to artificial marking difficult, the problem that the fiber attribute information is not accurate enough is extracted, the present application proposes a kind of image enhancement method with line Gauss convolution as core.Firstly, H-alpha image is normalized to preliminary stretch image contrast, while suppressing noise using Laplace-Gauss operator to enhance edge, then using line Gauss filter to enhance the contrast of solar filament fiber.In post-processing stage, using the adaptive histogram equalization of restriction contrast and top hat, bottom hat transformation method to improve the uneven problem of line Gauss filter enhancement effect.Finally, line Gauss filter is used again to further improve the contrast of solar filament fiber.After the enhancement of the method, the filamentary structure of solar filament fiber is clear, which makes it possible to objectively and accurately measure the filamentary structure attribute information of solar filament.

Description

Technical Field

[0001] This invention relates to a method for enhancing the fine structure of solar dark stripes based on linear Gaussian filtering, belonging to the field of image enhancement. Background Technology

[0002] Image enhancement is an important research topic in digital image processing, and it is widely used in fields such as remote sensing and astronomical imaging. Its main function is to purposefully emphasize the overall or local characteristics of an image, make an originally unclear image clearer or emphasize certain features of interest, amplify the differences between features of different objects in the image, suppress features of no interest, thereby improving image quality, enriching information, enhancing image interpretation and recognition, and meeting the needs of certain special analyses.

[0003] Dark filaments are the basic units that make up solar dark filaments, and these fibers depict the basic structure of the filament magnetic field. Dark filaments have very low contrast with the background and often appear densely in bundles, making it difficult to directly observe and trace the evolution of the fine filamentary structure of dark filaments using H-alpha high-resolution images. Image enhancement of the filamentary structure of dark filaments is beneficial to improving the scientific value of H-alpha spectral data on the fine structure of solar dark filaments. It is of great significance for further exploring the observational potential of existing ground-based solar telescopes, helping solar physicists to delve into unresolved issues such as the formation, evolution, and eruption models and mechanisms of the fine magnetic field structure of dark filaments, and promoting observational research on solar activity.

[0004] This invention proposes an image enhancement algorithm based on linear Gaussian filtering, which can significantly enhance the fine structure of solar dark stripes in H-alpha images, making the dark stripe fibers clearly distinguishable from the background, and helping scholars to observe dark stripe fibers and extract dark stripe fiber information more objectively. Summary of the Invention

[0005] This invention aims to use an image enhancement method based on linear Gaussian filtering to significantly enhance the intensity values ​​of solar dark stripe fiber pixels, which is characterized by low contrast in H-alpha images. This emphasizes the fiber structure information in the image, expands the distinction between the filamentous structure of solar dark stripes and the background, and solves the current problems of difficulty in manually marking dark stripe fibers and inaccurate acquisition of the position, length, direction and other attribute information of dark stripe fibers in H-alpha images.

[0006] The technical solution involved in this invention is: a method for enhancing the fine structure of solar filaments based on linear Gaussians, comprising the following steps:

[0007] Step 1: Preprocessing: Input H-alpha image, normalize it and perform Laplacian-Gaussian image sharpening to initially highlight the filamentary structure and suppress noise;

[0008] Step 1.1: Normalize the input H-alpha image. The pixel intensity values ​​in the H-alpha FITS file have a large distribution range (the difference between the maximum and minimum values ​​exceeds 30,000), resulting in low contrast between dark and background areas. Generally, the intensity histogram of an H-alpha image approximates a normal distribution. It can be normalized using the 3-sigma principle to improve the contrast between dark areas (such as dark bars and chromosomal fibers) and light areas in the image.

[0009]

[0010] Among them, I hα (x,y) represents the pixel intensity at position (x,y) in the H-alpha image; t low =max(I min ,I m -3σ hα ), t high =min(I max ,I m +3σ hα );I min I max I m and σ hα These represent the minimum, maximum, average, and standard deviation of the H-alpha image intensity, respectively. Furthermore, for ease of observation and analysis, we flip the normalized image to obtain image I, i.e., flipping light-colored pixels to dark-colored pixels. Equation (1) essentially increases the dynamic range of the dark and light regions in the H-alpha image and normalizes the intensity values ​​to [0,1].

[0011] Step 1.2: To further highlight the filamentary structure and suppress noise, the normalized image I is sharpened using the Laplacian-Gaussian method:

[0012]

[0013] In this equation, the Laplacian differential operator is used to sharpen the image within the right-hand side {}, and the regions with constant intensity are set to 0. The intensity values ​​of these constant regions can be recovered by superimposing this value with I. c is the sharpening intensity coefficient. G2(σ) represents a convolution operation, and G2(σ) represents a two-dimensional Gaussian filter with zero mean and standard deviation σ. Since the Laplacian operator is a second-order differential operator, it is very sensitive to small changes in the image. The purpose of convolving with the two-dimensional Gaussian function is to reduce the over-sharpening of small changes.

[0014] Step 2: Image enhancement of filamentous structures based on line Gaussian filters: Each pixel in the image and its neighborhood are rotated and convolved with a line Gaussian filter. The maximum value of the rotational convolution is the corresponding pixel value in the output image of this step, thereby enhancing the contrast of the filamentous fiber structure.

[0015] Accumulating intensity values ​​along the direction of the filamentous structure can improve the contrast of the dark filamentous structure. Therefore, the image I after sharpening according to equation (2) is... LoG Pixels distributed along different directions are convolved with a Gaussian filter to enhance contrast. That is,

[0016]

[0017] Where L(x,y,θ) represents the length of the Gaussian filter G. l (σ l The same as image I, with a width of one pixel and its center located in image I. LoG G is the line segment at position (x, y) that makes an angle θ with the bottom edge of the image. l (σ l ) is length This indicates rounding down, with the filter length determined according to the 3-sigma principle of a normal distribution (adding 1 ensures the filter is symmetric about the center), a width of one pixel, and a standard deviation of σ. l A linear Gaussian filter. For image I LoG For each pixel, line segment L is sampled by rotating around that pixel. Each sampling result is compared with the line using a Gaussian filter G. l (σ l Convolution, with rotation and sampling, yields the maximum value after convolution, which is the intensity value of the corresponding pixel in the enhanced image. For pixels on the fibrous structure, sampling along the fiber direction and convolving with a line Gaussian is essentially a Gaussian-weighted accumulation of intensity values ​​along the fiber direction. However, for background pixels outside the fibrous structure, the weighted accumulation of the maximum value does not significantly change their intensity value. Therefore, this process exhibits significant direction selectivity, increasing the intensity difference between the fibrous structure and the background, thus achieving fibrous structure image enhancement.

[0018] Step 3, Post-processing: Adaptive histogram equalization with contrast limitation and a combination of top-hat and low-hat transformations are used to suppress the increase in intensity values ​​of background pixels.

[0019] Step 3.1: To address the issue of uneven intensity after linear Gaussian filtering, Contrast Limited Adaptive Histogram Equalization (CLAHE) is employed. Compared to traditional histogram equalization, CLAHE effectively improves the local contrast of different regions of the image by calculating the Cumulative Distribution Function (CDF) of each pixel's neighborhood. Furthermore, CLAHE effectively avoids excessive amplification of noise in locally flat areas by limiting the histogram height.

[0020] Step 3.2: To further improve the clarity of the filamentous fiber structure outline in the CLAHE results, the image output from Step 3.1 is further processed using grayscale image morphological top-hat and bottom-hat transformations.

[0021] I tb =I c +T hat (I c )-B hat (I c (4)

[0022] Among them, I c T represents the image after CLAHE processing. hat (I c ) and B hat (I c ) respectively represent the images I c Perform top-hat and bottom-hat transformations. Equation (4) makes the local maxima in the image larger by putting on a "top hat" and the local minima smaller by putting on a "bottom hat", thereby effectively improving the clarity of the filamentous structure.

[0023] Step 4: Enhance the image again using the linear Gaussian filter: further improve the contrast of the dark fibers.

[0024] Step 4 is the same as Step 2, but the input image for this step is the result of Step 3.2. Analysis of equation (3) shows that, relative to the length... The filamentous fiber structure allows for greater length enhancement through linear Gaussian convolution. The pixel intensity values ​​on the filamentous fiber structure. In order to highlight the longer filamentous fiber structure, the result of Step 3 is again subjected to linear Gaussian convolution using Equation (3) to achieve the final image enhancement of the solar dark stripe filamentous fiber structure.

[0025] The beneficial effects of this invention are:

[0026] (1) This invention proposes an unsupervised image enhancement method for solar dark filament structure based on line Gaussian filtering. This method does not require a large amount of training sample data. Compared with the H-alpha image before enhancement, the dark filament structure after enhancement is clearly visible, making it possible to objectively and accurately measure the attribute information of the dark filament structure.

[0027] (2) This method is not only applicable to enhancing dark filamentous structures or chromospheric fibers in H-alpha images, but also has reference value for enhancing images containing filamentous structures such as needle-like bodies and coronal rings. Attached Figure Description

[0028] Figure 1 This is a flowchart of the present invention;

[0029] Figure 2 These are processing effect diagrams for each step in the embodiments of the present invention. Detailed Implementation

[0030] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0031] Example 1, as Figure 1 As shown, a method for enhancing the fine structure of solar filaments based on linear Gaussian filtering includes the following steps:

[0032] Step 1.1: Input an H-alpha image with a resolution of 249x375 (see...) Figure 2 (1) The maximum pixel intensity of this image is 64743 and the minimum is 25527, a difference of 39216. The huge difference in pixel intensity will result in low contrast between dark areas and background areas in the image. Normalization can be performed using the 3 sigma principle to improve the contrast between dark areas such as dark bars and color sphere fibers and light areas in the image:

[0033]

[0034] Among them, I hα (x,y) represents the pixel intensity at position (x,y) in the H-alpha image; t low =max(I min ,I m -3σ hα ), t high =min(I max ,I m +3σ hα );I min I max I m and σ hαThese represent the minimum, maximum, average, and standard deviation of the H-alpha image intensity, respectively. Furthermore, for ease of observation and analysis, we flipped the normalized image, i.e., flipped light-colored pixels to dark-colored pixels. In subsequent descriptions, I refers to the normalized and flipped image. Equation (1) essentially increases the dynamic range of the H-alpha image, including both dark and light regions, and normalizes the intensity values ​​to [0,1]. The processing effect of Step 1.1 is shown in [see...]. Figure 2 (2).

[0035] Step 1.2: To further highlight the filamentary structure and suppress noise, the normalized image I is sharpened and enhanced using the Laplacian-Gaussian method.

[0036]

[0037] In the equation, the Laplacian differential operator is used to sharpen the image within the square brackets {}. The Laplacian differential operator is a commonly used method for image edge extraction, and it sets the intensity of constant regions in the image to 0. By superimposing it with I, the intensity value of these constant regions can be recovered. c is the sharpening intensity coefficient; the larger c is, the more pronounced the sharpening of image edges. Here, c = 5. G2(σ) represents the convolution operation, and G2(σ) represents a two-dimensional Gaussian filter with zero mean and standard deviation σ. In this embodiment, σ = 1. Since the Laplacian operator is a second-order differential operator, it is very sensitive to small changes in the image. The purpose of convolving with the two-dimensional Gaussian function is to reduce over-sharpening of small changes. Additionally, the two-dimensional Gaussian filter can effectively suppress Poisson noise, which is common in astronomical images. The processing results of Step 1.2 are shown in [link to Step 1.2]. Figure 2 (3).

[0038] Step 2: Accumulating intensity values ​​along the direction of the filamentous structure can improve the contrast of the dark filamentous structure. Therefore, the image I after sharpening according to equation (2) is... LoG Pixels distributed along different directions are convolved with a Gaussian filter to achieve [improvement / enhancement]. The purpose is to increase contrast. That is,

[0039]

[0040] Where L(x,y,θ) represents the length of the Gaussian filter G. l (σ l The same as image I, with a width of one pixel and its center located in image I. LoG G is the line segment at position (x, y) that makes an angle θ with the bottom edge of the image. l (σ l ) is length (This indicates rounding down; the filter length is determined according to the 3-sigma principle of normal distribution; adding 1 ensures the filter is symmetric about the center.) The width is a single pixel, and the standard deviation is σ.l A linear Gaussian filter. In this embodiment, a σ-band filter is selected. l =5, a Gaussian line filter with a length of 31 pixels. For image I LoG For each pixel, line segment L is sampled by rotating around that pixel. Each sampling result is compared with the line using a Gaussian filter G. l (σ l Convolution, rotation sampling, and the maximum value after convolution are the intensity values ​​of the corresponding pixels in the enhanced image. For pixels on the fibrous structure, sampling the pixel value along the fiber direction and convolving it with a line Gaussian is essentially accumulating the intensity value with Gaussian weights along the fiber direction; however, for background pixels outside the fibrous structure, the maximum value of the weighted accumulation does not significantly change their intensity value. Therefore, this process has obvious direction selectivity, which can increase the intensity difference between the fibrous structure and the background, thus achieving fibrous structure image enhancement. The processing results of Step 2 are shown in... Figure 2 (4).

[0041] Step 3.1: To address the issue of uneven intensity after Gaussian filtering, a contrast-limited adaptive histogram equalization method (CLAHE) is employed. Compared to traditional histogram equalization, CLAHE effectively improves the local contrast of different regions of the image by calculating the cumulative distribution function (CDF) of each pixel's neighborhood. Furthermore, CLAHE effectively avoids excessive amplification of noise in locally flat areas by limiting the histogram height. In this embodiment, the image is divided into small blocks of equal size, and histogram equalization is applied to each block. The length and width of each block are 1 / 5 of the original image's length and width, respectively. To control the dynamic range, the number of bins in the histogram is set to 1024.

[0042] Step 3.2: To further improve the clarity of the filamentous fiber structure outline in the CLAHE results, the image output from Step 3.1 is further processed using grayscale image morphological top-hat and bottom-hat transformations.

[0043] I tb =I c +T hat (I c )-B hat (I c (4)

[0044] Among them, I c T represents the image after CLAHE processing. hat (I c ) and B hat (I c) represent the top-hat and bottom-hat transformations of the image, respectively. In this embodiment, a square with a width of 2 pixels is used as the structuring element. Equation (4) makes the local maxima in the image larger by wearing a "top hat" and the local minima smaller by wearing a "bottom hat", thereby effectively improving the clarity of the filamentous structure. The processing results of Step 3 are shown in Figure 2 (5).

[0045] Step 4 is the same as Step 2, but the input image for this step is the result of Step 3.2. Analysis of equation (3) shows that, relative to the length... The filamentous fiber structure allows for greater length enhancement through linear Gaussian convolution. The pixel intensity values ​​on the filamentous fiber structure. To highlight the longer filamentous fiber structure, the result of Step 3 is again subjected to linear Gaussian convolution using Equation (3) to achieve the final image enhancement of the solar dark stripe filamentous fiber structure. The processing effect of Step 4 is shown in the figure. Figure 2 (6) Its effect after black and white flip is as follows Figure 2 (7).

Claims

1. A method for enhancing the fine structure of solar fringe based on linear Gaussian filtering, characterized in that, Includes the following steps: Step 1: Preprocessing: Input H-alpha image, normalize it and perform Laplacian-Gaussian image sharpening to initially highlight the filamentary structure and suppress noise; Step 2: Image enhancement of filamentous structures based on line Gaussian filters: Each pixel in the image and its neighborhood are rotated and convolved with a line Gaussian filter. The maximum value of the rotational convolution is the corresponding pixel value in the output image of this step, thereby enhancing the contrast of the filamentous fiber structure. Step 3, Post-processing: Adaptive histogram equalization with contrast limitation and a combination of top-hat and low-hat transformations are used to suppress the increase in intensity values ​​of background pixels. Step 4: Enhance the image again using a line Gaussian filter: further improve the contrast of the dark striped filamentous fiber structure; The specific steps of Step 1 are as follows: Step 1.1: Normalize the input H-alpha image using the 3-sigma principle and enhance the contrast between dark and light regions in the image. （1） in, Indicates the H-alpha image ( x , y Pixel intensity at position ) , ; , , and The minimum, maximum, average, and standard deviation of the H-alpha image intensity are given. The normalized image is then flipped to obtain the image. I This means that light-colored pixels are flipped to dark-colored pixels; Step 1.2, on the image I Sharpening is performed using the Laplacian-Gaussian method: （2） In this equation, the Laplacian differential operator is used to sharpen the image within the right-hand side {}, and the regions of constant intensity in the image are set to 0. This is achieved by comparing the image with... I Superimposed, the intensity value of the constant region can be recovered. c This is the sharpening intensity factor. This represents the convolution operation. G 2( σ () indicates a mean of 0 and a standard deviation of 0. σ A two-dimensional Gaussian filter; The specific steps of Step 2 are as follows: Sharpened image I LoG Pixels distributed along different directions are convolved with a Gaussian filter to enhance contrast, i.e., （3） in, Represents the length and line Gaussian filter Same, width is one pixel, center located in the image I LoG At point (x, y), the angle between the point and the bottom edge of the image is... θ line segments, It is length Width is one pixel, standard deviation is A linear Gaussian filter for the image I LoG Each pixel, line segment L Rotate sampling around this pixel, and each sampling result is compared with a Gaussian filter. Convolution, the maximum value after rotation sampling convolution is the intensity value of the corresponding pixel in the enhanced image.

2. The method for enhancing the fine structure of solar fringe based on linear Gaussian filtering according to claim 1, characterized in that: The specific steps of Step 3 are as follows: Step 3.1: Adaptive histogram equalization with limited contrast is adopted. By calculating the cumulative distribution function of each pixel's neighborhood, the local contrast of each region of the image is improved, while avoiding excessive amplification of noise in local flat areas. Step 3.2: Apply grayscale image morphological top-hat and bottom-hat transformations to the image output in Step 3.

1. Further processing: （4） in, and Representing the images respectively Perform top-hat and bottom-hat transformations.

3. The method for enhancing the fine structure of solar fringe based on linear Gaussian filtering according to claim 1, characterized in that: Step 4 uses the same linear Gaussian filter as Step 2 to perform linear Gaussian convolution on the image processed in Step 3, thereby enhancing the image of the solar dark stripe filamentous structure.

Images