Method for removing image stripe noise and electronic device

Abstract

This disclosure provides a method and electronic device for removing stripe noise from an image. The method includes: acquiring original image data and converting it into a luminance-grayscale image; determining the stripe direction; performing a frequency domain transformation on the luminance-grayscale image along the stripe direction to obtain a one-dimensional frequency domain signal, and determining the stripe characteristic frequency and sidelobe frequency range; constructing a one-dimensional notch filter and performing an inverse Fourier transform to obtain an initial time-domain filter coefficient sequence; performing threshold filtering and normalization on the initial time-domain filter coefficient sequence based on the statistical variance of the initial time-domain filter coefficient sequence to obtain a one-dimensional time-domain filter kernel with a length less than the length of the original sequence; and performing a convolution operation on each row or column of pixel sequences in the luminance-grayscale image perpendicular to the stripe direction with the one-dimensional time-domain filter kernel to obtain a filtered pixel sequence and thus a processed image with stripe noise removed.

Description

Technical Field

[0001] This disclosure relates to the technical fields of image processing, and in particular to a method, electronic device, storage medium, and program product for removing image stripe noise. Background Technology

[0002] In practical applications of medical endoscopes, due to factors such as the circuit design characteristics of the sensor itself, power supply fluctuations of the endoscope equipment, unstable light intensity of the probe, and complex light reflection and electromagnetic interference inside the human body cavity, the output image is prone to periodic stripe noise (mostly alternating bright and dark stripes in the horizontal or vertical direction).

[0003] Such stripe interference can obscure the details of lesions in endoscopic images, making it difficult for doctors to accurately identify the location and shape of lesions, and seriously affecting the reliability of medical diagnosis.

[0004] Currently, techniques for removing stripes from endoscopic images are mainly divided into two categories: time-domain filtering and frequency-domain filtering. Time-domain filtering techniques (such as median filtering, mean filtering, and adaptive filtering) suppress stripes by directly performing neighborhood operations on image pixels. Their advantages are low computational cost and compatibility with embedded hardware, but they are difficult to accurately distinguish between stripe interference and image details, easily leading to image blurring and loss of details. Frequency-domain filtering techniques (such as notch filtering and low-pass filtering) convert the image to the frequency domain, locate and suppress the frequency components corresponding to the stripes, and then convert it back to the time domain to obtain a stripe-free image. Their advantages are accurate stripe removal and high image fidelity, but traditional frequency-domain filtering often uses two-dimensional filters, which result in a large filter kernel size after conversion to the time domain, leading to high computational complexity and difficulty in meeting real-time requirements. Summary of the Invention

[0005] This disclosure provides a method, electronic device, storage medium, and program product for removing image stripe noise.

[0006] According to one aspect of this disclosure, a method for removing stripe noise from an image is provided, comprising: Acquire raw image data and convert it into a brightness-grayscale image; Calculate the pixel grayscale variance in the horizontal and vertical directions of the brightness grayscale image, and determine the stripe direction based on the magnitude of the pixel grayscale variance; The brightness grayscale image is transformed in the frequency domain along the stripe extension direction to obtain a one-dimensional frequency domain signal, and the stripe characteristic frequency and the sidelobe frequency range of the stripe characteristic frequency are determined by the amplitude spectrum analysis of the one-dimensional frequency domain signal. Based on the fringe characteristic frequency and the sidelobe frequency range, a one-dimensional notch filter is constructed in the frequency domain corresponding to the one-dimensional frequency domain signal, and the one-dimensional notch filter is subjected to inverse Fourier transform to obtain an initial time-domain filter coefficient sequence. The initial time-domain filter coefficient sequence is subjected to threshold filtering and normalization based on the statistical variance of the initial time-domain filter coefficient sequence to obtain a one-dimensional time-domain filter kernel with a length shorter than the original sequence length of the initial time-domain filter coefficient sequence. For each row or column of pixels in the brightness grayscale image perpendicular to the stripe direction, a convolution operation is performed sequentially with the one-dimensional temporal filtering kernel to obtain the filtered pixel sequence, thereby obtaining the processed image with stripe noise removed.

[0007] According to at least one embodiment of the method for removing image stripe noise, the step of acquiring original image data and converting it into a brightness grayscale image includes: Determine the format of the acquired raw image data; If it is in Bayer format, the complete red, green and blue three-channel color information of each pixel position is first restored by the de-mosaic interpolation algorithm to obtain the RGB image; If it is in RGB format, then directly obtain the RGB image; The luminance value of each pixel in the RGB image is calculated using the weighted formula Gray=0.299×R+0.587×G+0.114×B, and the RGB image is converted into the luminance grayscale image.

[0008] According to at least one embodiment of the present disclosure, a method for removing image stripe noise, wherein calculating the pixel gray-level variance in the horizontal and vertical directions of the brightness gray-level image, and determining the stripe direction based on the magnitude of the pixel gray-level variance, includes: The grayscale variance of each row of pixels in the brightness grayscale image is calculated to obtain the horizontal variance sequence, and the grayscale variance of each column of pixels is calculated to obtain the vertical variance sequence. Calculate the average value of the horizontal variance sequence as the overall horizontal variance, and calculate the average value of the vertical variance sequence as the overall vertical variance. Compare the overall variance in the horizontal direction with the overall variance in the vertical direction; If the overall variance in the vertical direction is greater than the overall variance in the horizontal direction and the difference exceeds a preset judgment threshold, then the stripe direction is determined to be horizontal. If the overall variance in the horizontal direction is greater than the overall variance in the vertical direction and the difference exceeds the judgment threshold, then the stripe direction is determined to be vertical.

[0009] According to at least one embodiment of the present disclosure, a method for removing image stripe noise, wherein performing a frequency domain transformation on the brightness grayscale image along the stripe extension direction to obtain a one-dimensional frequency domain signal includes: When the direction of the stripes is determined to be horizontal: A two-dimensional Fourier transform is performed on the brightness grayscale image to obtain a two-dimensional frequency domain image; the data of each row of the two-dimensional frequency domain image are accumulated or averaged to obtain the one-dimensional frequency domain signal. Alternatively, the pixels of each row of the brightness grayscale image are accumulated to obtain a one-dimensional spatial sequence; a one-dimensional Fourier transform is performed on the one-dimensional spatial sequence to obtain the one-dimensional frequency domain signal. When the direction of the stripes is determined to be vertical: A two-dimensional Fourier transform is performed on the brightness grayscale image to obtain a two-dimensional frequency domain image; the data in each column of the two-dimensional frequency domain image are accumulated or averaged to obtain the one-dimensional frequency domain signal. Alternatively, the pixels in each column of the brightness grayscale image are accumulated to obtain a one-dimensional spatial sequence; a one-dimensional Fourier transform is performed on the one-dimensional spatial sequence to obtain the one-dimensional frequency domain signal.

[0010] A method for removing image stripe noise according to at least one embodiment of the present disclosure, wherein determining the stripe feature frequency and the sidelobe frequency range of the stripe feature frequency by amplitude spectrum analysis of the one-dimensional frequency domain signal includes: Calculate the amplitude spectrum of the one-dimensional frequency domain signal; Set the DC component corresponding to the zero-frequency position in the amplitude spectrum to zero; Find the peak with the largest amplitude in the amplitude spectrum after setting it to zero, and extract the frequency value corresponding to the peak as the fringe feature frequency. The sidelobe width is determined based on the half-width at half maximum (FWHM) of the spectral peak with the maximum amplitude, and the sidelobe frequency range is defined as: [fringe characteristic frequency − sidelobe width, fringe characteristic frequency + sidelobe width], with the fringe characteristic frequency as the center frequency.

[0011] A method for removing image stripe noise according to at least one embodiment of the present disclosure, wherein constructing a one-dimensional notch filter in the frequency domain corresponding to the one-dimensional frequency domain signal based on the stripe feature frequency and the sidelobe frequency range includes: The one-dimensional notch filter is constructed based on the Gaussian function, and its frequency domain expression is as follows: H(f)=1−exp(−(f−f0)² / (2σ²)); Where f0 is the fringe feature frequency, and σ is the Gaussian kernel width; The Gaussian kernel width σ is adaptively calculated based on the sidelobe width, and the calculation formula is as follows: σ = Δf / 1.177, where Δf is the sidelobe width; This makes the gain of the one-dimensional notch filter approach 0 in the sidelobe frequency range and approach 1 outside the sidelobe frequency range.

[0012] According to at least one embodiment of the method for removing image stripe noise, the step of performing threshold filtering and normalization processing on the initial temporal filter coefficient sequence based on the statistical variance of the initial temporal filter coefficient sequence to obtain a one-dimensional temporal filter kernel with a length smaller than the original sequence length of the initial temporal filter coefficient sequence includes: Calculate the overall statistical variance of the initial time-domain filter coefficient sequence; The screening threshold is dynamically determined based on the overall statistical variance, wherein the larger the overall statistical variance, the larger the screening threshold is set to. Starting from the center of the initial time-domain filter coefficient sequence, the system iterates towards both ends to select coefficients whose absolute value is greater than the selection threshold, thus forming an effective coefficient sequence. The effective coefficient sequence is normalized so that the sum of all coefficients is 1, and the normalized effective coefficient sequence is used as the one-dimensional time-domain filter kernel.

[0013] According to at least one embodiment of the method for removing image stripe noise, the process of traversing and filtering from the center position of the initial temporal filter coefficient sequence to both ends further includes at least one of the following termination conditions: When the absolute value of a preset number of coefficients does not exceed the filtering threshold for the first time in any one end, further traversal to that end is stopped, and all subsequent coefficients of that end are discarded. Alternatively, when the number of selected coefficients reaches the preset maximum filter kernel length, the filtering stops, and the currently selected coefficients form the effective coefficient sequence.

[0014] According to at least one embodiment of the present disclosure, a method for removing image stripe noise includes, in which each row or column of pixel sequences in the brightness grayscale image perpendicular to the stripe direction is sequentially convolved with the one-dimensional temporal filter kernel, comprising: When the stripe direction is determined to be horizontal, the convolution operation is performed on each column of pixel sequence of the brightness grayscale image with the one-dimensional temporal filter kernel. When the direction of the stripes is determined to be vertical, the convolution operation is performed on each row of pixel sequence of the brightness grayscale image with the one-dimensional temporal filter kernel.

[0015] A method for removing image stripe noise according to at least one embodiment of the present disclosure, the method being applied to medical endoscope image processing, further comprising, after obtaining the processed image with stripe noise removed: The processed image is then output to a display device for display.

[0016] According to another aspect of this disclosure, an electronic device is provided, comprising: a memory storing execution instructions; and a processor executing the execution instructions stored in the memory, causing the processor to perform a method for removing image stripe noise according to any embodiment of this disclosure.

[0017] According to another aspect of this disclosure, a readable storage medium is provided that stores executable instructions, which, when executed by a processor, are used to implement a method for removing image stripe noise according to any embodiment of this disclosure.

[0018] According to another aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements a method for removing image stripe noise according to any embodiment of this disclosure. Attached Figure Description

[0019] The accompanying drawings illustrate exemplary embodiments of the present disclosure and, together with the description thereof, serve to explain the principles of the present disclosure. These drawings are included to provide a further understanding of the present disclosure and are incorporated in and constitute a part of this specification.

[0020] Figure 1 This is a flowchart of a method for removing image stripe noise according to one embodiment of the present disclosure.

[0021] Figure 2 This is a flowchart illustrating the determination of stripe direction in a method for removing image stripe noise according to one embodiment of the present disclosure.

[0022] Figure 3 This is a schematic flowchart illustrating the process of determining the fringe feature frequency and the sidelobe frequency range of the fringe feature frequency in a method for removing image fringe noise according to one embodiment of the present disclosure.

[0023] Figure 4 This is a schematic flowchart illustrating the process of obtaining a one-dimensional temporal filter kernel in a method for removing image stripe noise according to one embodiment of the present disclosure.

[0024] Figure 5 This is a schematic block diagram of an electronic device according to one embodiment of the present disclosure. Detailed Implementation

[0025] The present disclosure will now be described in further detail with reference to the accompanying drawings and examples. It should be understood that the specific examples described herein are for illustrative purposes only and are not intended to limit the scope of the disclosure. Furthermore, it should be noted that, for ease of description, only the parts relevant to the present disclosure are shown in the accompanying drawings.

[0026] It should be noted that, where there is no conflict, the embodiments and features described in this disclosure can be combined with each other. The technical solutions of this disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0027] Figure 1 This is a flowchart of a method for removing image stripe noise according to one embodiment of the present disclosure.

[0028] refer to Figure 1 In some embodiments of this disclosure, the method M100 for removing image stripe noise includes steps S102 to S112.

[0029] S102: Acquire raw image data and convert it into a brightness grayscale image.

[0030] Medical endoscopic images are typically color images, and their stripe noise (whether horizontal or vertical stripes) manifests as brightness fluctuations in terms of light and dark variations. Converting a color image to a grayscale image extracts the light and dark variation information of the stripes into a single channel, facilitating subsequent operations on the brightness signal, such as variance statistics and frequency domain transformation.

[0031] The raw data output by the endoscope image sensor is mainly in two formats: Bayer format (Bayer array) and RGB format.

[0032] Bayer format is the sensor's raw acquisition format, where each pixel location records only one of the three colors: red (R), green (G), and blue (B). RGB format, on the other hand, provides each pixel with complete R, G, and B channel values.

[0033] In this step, the format of the acquired raw image data is first determined.

[0034] If it is Bayer format, the first step is to use a demosaic interpolation algorithm, for example, to reconstruct the missing values ​​of the other two colors for each pixel position by interpolating the known color information of adjacent pixels, thereby restoring the complete red, green and blue three-channel color information for each pixel position and obtaining an RGB image.

[0035] If it is in RGB format, then directly obtain the RGB image.

[0036] Then, for each pixel of the RGB image, the brightness value can be calculated using a weighted formula: Gray = 0.299 × R + 0.587 × G + 0.114 × B; Iterate through all pixels in the image and convert the RGB image to a grayscale image.

[0037] S104. Calculate the pixel grayscale variance in the horizontal and vertical directions of the brightness grayscale image, and determine the stripe direction based on the magnitude of the pixel grayscale variance.

[0038] Stripe noise has a clear directionality. For horizontal stripes, the extension direction is horizontal, while the changes in brightness are reflected in the vertical direction, resulting in significant differences in pixel grayscale in the vertical direction; for vertical stripes, the extension direction is vertical, while the changes in brightness are reflected in the horizontal direction, resulting in significant differences in pixel grayscale in the horizontal direction.

[0039] Specifically, firstly, the grayscale variance of each row of pixels in the above brightness grayscale image is calculated to obtain the horizontal variance sequence, and then the grayscale variance of each column of pixels is calculated to obtain the vertical variance sequence.

[0040] "Variance" refers to the degree of dispersion of the grayscale value of each pixel in a row or column from the average grayscale value of that row or column. The larger the variance, the more drastic the fluctuation of the pixel grayscale value in that row or column. The presence of periodic stripe noise will cause periodic bright-dark transitions in the pixel grayscale value perpendicular to its extension direction, thus generating a large variance in that direction.

[0041] To eliminate the interference of abnormal variance fluctuations that may occur in individual rows or columns on the overall judgment, the average value of all variances in the horizontal variance sequence is calculated as the overall horizontal variance representing the degree of gray-level change of the entire image in the horizontal direction. The average value of all variances in the vertical variance sequence is also calculated as the overall vertical variance representing the degree of gray-level change of the entire image in the vertical direction.

[0042] Finally, a preset judgment threshold is set, which can be set to a small positive number based on experience.

[0043] Next, we compare the overall variance in the horizontal direction with the overall variance in the vertical direction.

[0044] When the overall variance in the vertical direction is greater than the overall variance in the horizontal direction, and the difference between the two exceeds the judgment threshold, the stripe direction is determined to be horizontal, that is, the stripes extend in the horizontal direction, and the grayscale changes are mainly reflected in the vertical direction.

[0045] When the overall variance in the horizontal direction is greater than the overall variance in the vertical direction, and the difference between the two exceeds the judgment threshold, the stripe direction is determined to be the vertical direction, that is, the stripes extend along the vertical direction, and the grayscale changes are mainly reflected in the horizontal direction.

[0046] By setting a judgment threshold, it is possible to filter out minute variance differences caused by random noise or the image content itself, thus avoiding incorrect judgments when there are no obvious stripes or the stripes are extremely weak in the image.

[0047] S106. Perform frequency domain transformation on the brightness grayscale image along the extension direction of the stripes to obtain a one-dimensional frequency domain signal, and determine the stripe characteristic frequency and the sidelobe frequency range of the stripe characteristic frequency by analyzing the amplitude spectrum of the one-dimensional frequency domain signal.

[0048] Since the stripe noise output by the sensor is mostly periodic stripes in the horizontal or vertical direction, it manifests as discrete amplitude peaks in the frequency domain. This step utilizes this frequency domain characteristic to locate the frequency position and distribution range of the stripe noise in the frequency domain.

[0049] First, a frequency domain transformation is performed, and the two-dimensional image information is compressed into a one-dimensional frequency domain signal according to the fringe direction determined by S104.

[0050] The first approach is to first perform a two-dimensional Fourier transform on the brightness grayscale image Gray(x,y), and then accumulate or average the corresponding data of the two-dimensional frequency domain image along the direction of the stripe extension to compress it into a one-dimensional frequency domain signal.

[0051] The second approach is to first accumulate the corresponding pixel values ​​of the grayscale image along the direction of the stripe extension in the spatial domain to obtain a one-dimensional spatial sequence, and then perform a one-dimensional Fourier transform on the one-dimensional spatial sequence to obtain a one-dimensional frequency domain signal.

[0052] Specifically, the grayscale image Gray(x,y) can be transformed to the frequency domain using a two-dimensional Fourier transform. The frequency domain image is then averaged along the stripe direction to obtain one-dimensional frequency domain data. Alternatively, the grayscale image can be accumulated along the stripe direction first, and then a one-dimensional Fourier transform can be performed to obtain one-dimensional frequency domain data. .

[0053] For example, brightness and grayscale images The image size is There are vertical stripes; a two-dimensional Fourier transform can be performed first. The average of each column in the frequency domain image is calculated. ; You can also The average of each column in the image is calculated. Then perform a one-dimensional Fourier transform. ; This disclosure can concentrate the energy of the stripe signal by accumulating or averaging the data along the stripe extension direction, while averaging out some random noise and enhancing the signal-to-noise ratio in the frequency domain.

[0054] In the amplitude spectrum analysis stage, the amplitude spectrum of the one-dimensional frequency domain signal is first calculated. Since the zero frequency (f=0) corresponds to the DC component, which only represents the overall brightness level of the image and does not contain stripe information, this DC component is first set to zero to eliminate interference. Then, the peak with the largest amplitude is found in the zeroed amplitude spectrum. Since stripe noise is periodic, it manifests as a discrete peak with concentrated energy in the frequency domain. This largest peak corresponds to the dominant frequency component of the stripe noise, and its corresponding frequency value is extracted as the stripe characteristic frequency f0. At the same time, considering that the actual stripes are not ideal single-frequency sine waves, their frequency domain energy will expand to both sides with f0 as the center, forming a sidelobe frequency range. Therefore, this disclosure also defines the sidelobe frequency range with f0 as the center frequency and Δf as the radius.

[0055] The above processing effectively compresses two-dimensional image information into one-dimensional frequency domain data, which facilitates the accurate extraction of stripe feature frequencies and solves the technical problems of high computational complexity and difficulty in meeting real-time requirements of traditional two-dimensional filters.

[0056] S108. Based on the fringe characteristic frequency and the sidelobe frequency range, construct a one-dimensional notch filter in the frequency domain corresponding to the one-dimensional frequency domain signal, and perform an inverse Fourier transform on the one-dimensional notch filter to obtain an initial time-domain filter coefficient sequence.

[0057] Based on the fringe characteristic frequency and sidelobe frequency range determined in S106, this step constructs a one-dimensional notch filter in the frequency domain to suppress fringe noise. This filter is designed to suppress fringe noise within the frequency range corresponding to the fringe, while allowing the signal to pass outside this range.

[0058] After the one-dimensional notch filter is constructed, the frequency domain filter is transformed to the time domain through inverse Fourier transform to obtain the initial time domain filter coefficient sequence, which provides the coefficient sequence for subsequent time domain filtering operations.

[0059] S110. The initial time-domain filter coefficient sequence is subjected to threshold screening and normalization based on the statistical variance of the initial time-domain filter coefficient sequence to obtain a one-dimensional time-domain filter kernel with a length less than the original sequence length of the initial time-domain filter coefficient sequence.

[0060] Since the initial time-domain filter coefficient sequence is usually quite long, it contains a large number of coefficients that contribute little to the filtering effect. This step involves statistically analyzing the initial time-domain filter coefficient sequence, determining the screening criteria based on its statistical characteristics, removing coefficients that contribute little to the filtering effect, retaining coefficients that contribute significantly to the filtering effect, and then normalizing them to obtain a one-dimensional time-domain filter kernel with a significantly reduced length.

[0061] S112. For each row or column of pixel sequence in the brightness grayscale image that is perpendicular to the stripe direction, perform convolution operation with the one-dimensional temporal filtering kernel in sequence to obtain the filtered pixel sequence, so as to obtain the processed image with stripe noise removed.

[0062] Based on the previously determined stripe direction, this step performs convolution on each column of pixel sequence when the stripes are horizontal and on each row of pixel sequence when the stripes are vertical.

[0063] The directional convolution method disclosed herein enables the filtering operation to be accurately matched with the stripe extension direction, thereby efficiently removing periodic stripe noise while keeping the pixel values ​​in non-stripe directions unchanged, preserving the original image details to the greatest extent, thus obtaining a complete image after stripe noise removal.

[0064] The method for removing stripe noise from an image disclosed herein first converts the original image data into a brightness grayscale image to highlight the brightness variations of the stripes in the image. Next, the direction of the stripes is determined by comparing the pixel grayscale variances of the brightness grayscale images in two orthogonal directions. Then, frequency domain transformation and amplitude spectrum analysis are performed on the image along the stripe direction to obtain the characteristic frequencies and sidelobe frequency ranges of the stripe noise. Based on this, a one-dimensional notch filter is constructed in the frequency domain and then converted to the time domain via inverse Fourier transform. Subsequently, the time-domain coefficient sequence is subjected to thresholding and normalization based on statistical variance to obtain a small-sized one-dimensional time-domain filter kernel. Finally, one-dimensional convolution filtering is performed along the direction perpendicular to the stripes to obtain the processed image with stripe noise removed.

[0065] The method for removing image stripe noise disclosed herein accurately locates stripe frequency information through frequency domain analysis and designs filters accordingly. The filtering operation is then converted to the time domain and executed efficiently using a one-dimensional convolution method. This method effectively removes periodic stripe noise while preserving image details, significantly reducing computational complexity and meeting the real-time requirements of scenarios such as medical endoscopes. Figure 2 This is a flowchart illustrating the determination of stripe direction in a method for removing image stripe noise according to one embodiment of the present disclosure.

[0066] refer to Figure 2 In some embodiments of this disclosure, step S104 described above, which involves calculating the pixel grayscale variance in the horizontal and vertical directions of the brightness grayscale image and determining the stripe direction based on the magnitude of the pixel grayscale variance, includes: S1041. Calculate the grayscale variance of each row of pixels in the brightness grayscale image to obtain the horizontal variance sequence, and calculate the grayscale variance of each column of pixels to obtain the vertical variance sequence. S1042. Calculate the average value of the horizontal variance sequence as the overall horizontal variance, and calculate the average value of the vertical variance sequence as the overall vertical variance. S1043. Compare the overall variance in the horizontal direction with the overall variance in the vertical direction; S1044. If the overall variance in the vertical direction is greater than the overall variance in the horizontal direction and the difference exceeds a preset judgment threshold, then the stripe direction is determined to be horizontal. S1045. If the overall variance in the horizontal direction is greater than the overall variance in the vertical direction and the difference exceeds the judgment threshold, then the stripe direction is determined to be the vertical direction.

[0067] In this embodiment, in S1041, the grayscale variance of each row and each column of pixels in the brightness grayscale image is calculated respectively, resulting in two variance sequences. The variance measures the degree of dispersion of a pixel's grayscale value from its mean. The presence of periodic stripe noise causes periodic bright-dark transitions in pixel grayscale values ​​perpendicular to its extension direction, thus resulting in a larger variance in this direction.

[0068] In S1042, in order to eliminate the interference of local individual row or column abnormal fluctuations on the overall judgment, the average value of the two variance sequences is taken to obtain the overall variance representing the degree of gray-level change of the entire image in two directions.

[0069] In S1043 to S1045, the two overall variances are compared. When the overall variance in one direction is greater than that in another direction and the difference exceeds a preset threshold, the stripe direction is determined to be perpendicular to the direction with the larger variance, that is, the stripe extends along the direction with the smaller variance.

[0070] Setting a judgment threshold can filter out small variance differences caused by random noise or the image content itself, avoiding incorrect judgments when there are no obvious stripes or the stripes are very weak in the image, thus improving the robustness and accuracy of direction judgment.

[0071] In some embodiments of this disclosure, step S106, described above, of performing a frequency domain transformation on the brightness grayscale image along the stripe extension direction to obtain a one-dimensional frequency domain signal, includes: When the direction of the stripes is determined to be horizontal: A two-dimensional Fourier transform is performed on the brightness grayscale image to obtain a two-dimensional frequency domain image; the data of each row of the two-dimensional frequency domain image are accumulated or averaged to obtain the one-dimensional frequency domain signal. Alternatively, the pixels of each row of the brightness grayscale image are accumulated to obtain a one-dimensional spatial sequence; a one-dimensional Fourier transform is performed on the one-dimensional spatial sequence to obtain the one-dimensional frequency domain signal. When the direction of the stripes is determined to be vertical: A two-dimensional Fourier transform is performed on the brightness grayscale image to obtain a two-dimensional frequency domain image; the data in each column of the two-dimensional frequency domain image are accumulated or averaged to obtain the one-dimensional frequency domain signal. Alternatively, the pixels in each column of the brightness grayscale image are accumulated to obtain a one-dimensional spatial sequence; a one-dimensional Fourier transform is performed on the one-dimensional spatial sequence to obtain the one-dimensional frequency domain signal.

[0072] This implementation provides two paths for obtaining a one-dimensional frequency domain signal through frequency domain transformation.

[0073] In this embodiment, based on the determined fringe direction, it is possible to operate in the frequency domain (first perform a two-dimensional Fourier transform, and then average the data of each row or column along the fringe extension direction), or to operate in the spatial domain (first accumulate the pixel values ​​of each row or column along the fringe extension direction, and then perform a one-dimensional Fourier transform).

[0074] The two paths described above are mathematically equivalent and both can compress two-dimensional image information into a one-dimensional frequency domain signal. Accumulating or averaging along the fringe extension direction can concentrate the energy of the fringe signal, while averaging out some random noise, enhancing the signal-to-noise ratio in the frequency domain, and facilitating accurate extraction of fringe frequency features.

[0075] Figure 3 This is a schematic flowchart illustrating the process of determining the fringe feature frequency and the sidelobe frequency range of the fringe feature frequency in a method for removing image fringe noise according to one embodiment of the present disclosure.

[0076] refer to Figure 3 In some embodiments of this disclosure, step S106 described above, which involves determining the fringe characteristic frequency and the sidelobe frequency range of the fringe characteristic frequency through amplitude spectrum analysis of the one-dimensional frequency domain signal, includes: S1061. Calculate the amplitude spectrum of the one-dimensional frequency domain signal; S1062. Set the DC component corresponding to the zero-frequency position in the amplitude spectrum to zero; S1063. Find the spectral peak with the largest amplitude in the amplitude spectrum after setting it to zero, and extract the frequency value corresponding to the spectral peak as the fringe feature frequency. S1064. Determine the sidelobe width based on the half-width at half-maximum of the spectral peak, and define the sidelobe frequency range as: [fringe characteristic frequency − sidelobe width, fringe characteristic frequency + sidelobe width], with the fringe characteristic frequency as the center frequency.

[0077] In S1061, the amplitude spectrum of the one-dimensional frequency domain signal is first calculated.

[0078] In S1062, the DC component corresponding to the zero-frequency position in the amplitude spectrum is set to zero to eliminate the interference of the overall brightness level of the image on the frequency analysis.

[0079] In S1063, the peak with the largest amplitude is found in the amplitude spectrum after being zeroed. Since the stripe noise is periodic, it manifests as discrete peaks with concentrated energy in the frequency domain. Therefore, the largest peak corresponds to the dominant frequency component of the stripe noise, and its corresponding frequency value is extracted as the stripe characteristic frequency f0.

[0080] In S1064, considering that the actual stripes are not ideal single-frequency sine waves, their frequency domain energy will expand to both sides with f0 as the center to form side lobes. In order to accurately capture the complete frequency range of stripe noise, the half-width of the maximum spectral peak is measured as the side lobe width Δf, and with f0 as the center frequency, the side lobe frequency range is defined as [f0-Δf, f0+Δf].

[0081] Specifically, calculate the frequency domain signal amplitude spectrum Remove DC component ( (At this location, corresponding to the overall brightness of the image), the frequency corresponding to the maximum peak value in the amplitude spectrum is selected as the characteristic frequency of the stripe. Simultaneously record the sidelobe frequency range of the characteristic frequency f0. ,in This refers to the width of the side lobe.

[0082] The half-width at half-maximum (WHM) refers to the frequency range where the amplitude drops from the peak value to approximately 0.707 times the peak value, and Δf is typically 5% to 10% of f0.

[0083] In some embodiments of this disclosure, step S108 described above, which involves constructing a one-dimensional notch filter in the frequency domain corresponding to the one-dimensional frequency domain signal based on the fringe characteristic frequency and the sidelobe frequency range, includes: The one-dimensional notch filter is constructed based on the Gaussian function, and its frequency domain expression is as follows: H(f)=1−exp(−(f−f0)² / (2σ²)); Where f0 is the fringe feature frequency, and σ is the Gaussian kernel width; The Gaussian kernel width σ is adaptively calculated based on the sidelobe width, and the calculation formula is as follows: σ = Δf / 1.177, where Δf is the sidelobe width; This makes the gain of the one-dimensional notch filter approach 0 in the sidelobe frequency range and approach 1 outside the sidelobe frequency range.

[0084] In this embodiment, the Gaussian function is used as the basis function for the notch filter because the Gaussian function decays rapidly at the center frequency and has a smooth shape, making it suitable for constructing frequency domain notch filters.

[0085] The frequency domain expression of the filter is: Where f0 is the fringe characteristic frequency and σ is the Gaussian kernel width, which determines the width and shape of the notch.

[0086] In this disclosure, σ is not arbitrarily set, but is calculated based on the sidelobe width Δf using the formula σ=Δf / 1.177. This specific formula ensures that the constructed notch filter's suppression range precisely covers the sidelobe frequency range [f0-Δf, f0+Δf]. Within this range, the gain approaches 0 to completely suppress fringe noise. Outside the sidelobe frequency range, the filter gain rapidly recovers and approaches 1, allowing other normal frequency components in the image to pass through almost without loss. This design ensures thorough fringe suppression while avoiding an excessively wide suppression range that would suppress adjacent useful frequency components, thus preserving the original detail information of the image to the greatest extent possible.

[0087] Figure 4 This is a schematic flowchart illustrating the process of obtaining a one-dimensional temporal filter kernel in a method for removing image stripe noise according to one embodiment of the present disclosure.

[0088] refer to Figure 4 In some embodiments of this disclosure, step S110, described above, of performing threshold filtering and normalization processing on the initial time-domain filter coefficient sequence based on the statistical variance of the initial time-domain filter coefficient sequence to obtain a one-dimensional time-domain filter kernel with a length smaller than the original sequence length of the initial time-domain filter coefficient sequence, includes: S1101. Calculate the overall statistical variance of the initial time-domain filter coefficient sequence; S1102. Dynamically determine the screening threshold based on the overall statistical variance, wherein the larger the overall statistical variance, the larger the screening threshold is set to; S1103. Starting from the center of the initial time-domain filter coefficient sequence, simultaneously traverse to both ends to filter out coefficients whose absolute value is greater than the filtering threshold, thus forming an effective coefficient sequence. S1104. Normalize the effective coefficient sequence so that the sum of all coefficients is 1, and use the normalized effective coefficient sequence as the one-dimensional time-domain filter kernel.

[0089] This implementation provides a specific process for compressing and optimizing the initial time-domain filter coefficient sequence.

[0090] The initial time-domain filter coefficient sequence is expressed as follows: , The filter length is the same as the number of pixels in a row or column of the grayscale image.

[0091] In S1101, the overall statistical variance of the initial time-domain filter coefficient sequence is first calculated. This variance is used to measure the degree of dispersion of each coefficient value in the sequence from the mean of the sequence. The larger the variance, the higher the discrimination between the effective coefficients and the redundant coefficients.

[0092] In S1102, the screening threshold T is dynamically determined based on the variance. The principle is that the larger the variance, the larger the value of T, so as to more accurately remove redundant coefficients when the coefficient discrimination is high, and avoid accidentally deleting effective coefficients when the discrimination is low.

[0093] In S1103, the sequence is traversed from the center to both ends, and coefficients with absolute values ​​greater than the threshold T are selected to form an effective coefficient sequence. The center position corresponds to the main lobe of the filter kernel, which contains the coefficients that contribute the most to the filtering effect.

[0094] Due to the initial time-domain filter coefficient sequence In this process, there are a large number of redundant coefficients with extremely small absolute values, which increases the computational load. This step uses threshold filtering to remove redundant coefficients and generate a small-sized time-domain filter kernel. Setting the threshold T: The principle for setting a small threshold T is to eliminate redundant coefficients while retaining effective coefficients that contribute to stripe removal. T is adjusted based on the variance of the initial time-domain filter coefficient sequence; the larger the variance, the larger the value of T, and vice versa. This is because the initial time-domain filter coefficient sequence... The larger the variance, the more significant the difference between coefficients, and the higher the distinction between redundant coefficients and effective coefficients. Using a larger T can more accurately remove redundant coefficients while avoiding the accidental deletion of effective coefficients.

[0095] Coefficient selection: Extracting the initial time-domain filter coefficient sequence In the selection process, coefficients with absolute values ​​greater than a threshold T are discarded, while redundant coefficients with absolute values ​​less than or equal to T are removed, resulting in a filtered sequence of effective coefficients. Alternatively, the length of the effective coefficient sequence can be set directly, preserving the symmetrical position coefficients on the left and right sides of the center of the initial sequence, making them equal to the preset length.

[0096] Where M1 is the length of the effective coefficient sequence (i.e., the small-sized one-dimensional time-domain filter kernel after filtering), and M1 is usually much smaller than N1.

[0097] In S1104, the coefficients are normalized. Normalization process is performed to obtain This eliminates the impact of the coefficient magnitude difference on subsequent convolution operations.

[0098] The normalization formula can be: By summing all coefficients to 1, the potential impact of magnitude changes caused by coefficient selection on the overall image brightness is eliminated. After the above processing, the final filter kernel length is much smaller than the original sequence length, significantly reducing the computational cost of subsequent convolution operations.

[0099] In some embodiments of this disclosure, step S1103 described above, starting from the center of the initial time-domain filter coefficient sequence and traversing both ends simultaneously, filters out coefficients whose absolute value is greater than the filtering threshold to form an effective coefficient sequence, including at least one of the following termination conditions: When the absolute value of a preset number of coefficients does not exceed the filtering threshold for the first time in any one end, further traversal to that end is stopped, and all subsequent coefficients of that end are discarded. Alternatively, when the number of selected coefficients reaches the preset maximum filter kernel length, the filtering stops, and the currently selected coefficients form the effective coefficient sequence.

[0100] When traversing from the center of the sequence to both ends, the absolute value of the coefficients that are further away from the center is usually smaller and gradually approaches zero, thus contributing less to the filtering effect.

[0101] The first termination condition mentioned above is that when the absolute value of a preset number of coefficients in a certain direction does not exceed the screening threshold T for the first time, it indicates that the direction has entered the region with dense redundant coefficients. Continuing to traverse will unnecessarily increase the computational overhead, so the traversal of that direction is stopped and subsequent coefficients are discarded.

[0102] The second termination condition mentioned above is to stop filtering when the number of selected coefficients reaches the preset maximum filter kernel length. By setting the upper limit of the length in advance, the size of the final filter kernel is controlled to ensure that it meets the performance requirements of real-time processing.

[0103] The above termination condition, while ensuring the filtering effect, further improves the screening efficiency and avoids unnecessary traversal calculations.

[0104] In some embodiments of this disclosure, for each row or column of pixel sequence in the brightness grayscale image perpendicular to the stripe direction, as described above, convolution operations are performed sequentially with the one-dimensional temporal filter kernel, including: When the stripe direction is determined to be horizontal, the convolution operation is performed on each column of pixel sequence of the brightness grayscale image with the one-dimensional temporal filter kernel. When the direction of the stripes is determined to be vertical, the convolution operation is performed on each row of pixel sequence of the brightness grayscale image with the one-dimensional temporal filter kernel.

[0105] The above method, based on the determined fringe direction, determines the grayscale changes in a horizontal direction when the fringe extends horizontally, with the grayscale changes occurring in the vertical direction. Therefore, each column of pixel sequence in the brightness grayscale image is convolved with a one-dimensional temporal filter kernel. When the fringe extends vertically, the fringe extends vertically, with the grayscale changes occurring in the horizontal direction. Therefore, each row of pixel sequence in the brightness grayscale image is convolved with a one-dimensional temporal filter kernel.

[0106] The above method ensures that the direction of the filtering operation is consistent with the direction of gray-level change of the stripe noise, so that the effect of the filtering kernel is accurately applied to the brightness fluctuations caused by the stripes, thus preserving the original details of the image to the greatest extent.

[0107] The method M100 for removing image stripe noise according to the various embodiments described above, applied to medical endoscope image processing, further includes the following after obtaining the processed image with stripe noise removed: The processed image is then output to a display device for display.

[0108] In practical applications of medical endoscopes, due to factors such as the inherent circuit design characteristics of the sensors, power supply fluctuations, unstable probe light intensity, and complex light reflections and electromagnetic interference within the human body cavity, the images output by the endoscope are prone to exhibiting periodic alternating horizontal or vertical stripe noise. This stripe interference can obscure lesion details, making it difficult for doctors to accurately identify the location and shape of lesions. After obtaining the image with stripe noise removed through the aforementioned steps, the processed image is output to a display device (computer screen or monitor, etc.) for display, providing doctors with a clear and stable surgical field of view.

[0109] The execution subject of the method for removing image stripe noise in the specific embodiments of this disclosure can be an electronic device such as a computer.

[0110] Therefore, based on any of the above embodiments, this disclosure also provides an electronic device that can perform the method for removing image stripe noise according to any of the embodiments described above.

[0111] Figure 5 This is a schematic block diagram of an electronic device 1000 according to one embodiment of the present disclosure.

[0112] The hardware architecture of the electronic device 1000 can be implemented using a bus architecture. The bus architecture can include any number of interconnect buses and bridges, depending on the specific application of the hardware and overall design constraints. Bus 1100 connects various circuits, including one or more processors 1200, memory 1300, and / or hardware modules. Bus 1100 can also connect various other circuits 1400, such as peripheral devices, voltage regulators, power management circuits, external antennas, etc.

[0113] Bus 1100 can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, this diagram uses only one connection line, but this does not imply that there is only one bus or one type of bus.

[0114] This disclosure also provides a readable storage medium storing a computer program that, when executed by a processor, is used to implement the methods described above. A "readable storage medium" can be any means capable of containing, storing, communicating, propagating, or transmitting a program for use by or in conjunction with an instruction execution system, apparatus, or device. More specific examples of a readable storage medium include: an electrical connection with one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and portable read-only memory (CDROM), etc.

[0115] This disclosure also provides a computer program product, the methods of which can be implemented wholly or partially through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented wholly or partially as a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed, all or part of the processes or functions of this disclosure are performed.

[0116] Computer programs or instructions can be stored in a readable storage medium or transferred from one readable storage medium to another. For example, the computer program or instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The readable storage medium can be any available medium capable of access, or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium, such as a floppy disk, hard disk, or magnetic tape; an optical medium, such as a digital video optical disc; or a semiconductor medium, such as a solid-state drive. The computer-readable storage medium can be a volatile or non-volatile storage medium, or it can include both volatile and non-volatile types of storage media.

[0117] Those skilled in the art will understand that embodiments of this disclosure can be provided as methods, systems, or computer program products. Therefore, this disclosure can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this disclosure can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0118] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0119] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0120] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0121] In the description of this specification, the references to terms such as "one embodiment / mode," "some embodiments / modes," "example," "specific example," or "some examples," etc., refer to specific features, structures, or characteristics described in connection with that embodiment / mode or example, which are included in at least one embodiment / mode or example of this disclosure. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment / mode or example. Moreover, the specific features, structures, or characteristics described may be combined in any suitable manner in one or more embodiments / modes or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments / modes or examples described in this specification, as well as the features of different embodiments / modes or examples.

[0122] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this disclosure, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0123] Those skilled in the art should understand that the above embodiments are merely for illustrating the present disclosure and are not intended to limit the scope of the disclosure. Those skilled in the art can make other changes or modifications based on the above disclosure, and these changes or modifications still fall within the scope of the present disclosure.

Claims

1. A method for removing stripe noise from an image, characterized in that, include: Acquire raw image data and convert it into a brightness-grayscale image; Calculate the pixel grayscale variance in the horizontal and vertical directions of the brightness grayscale image, and determine the stripe direction based on the magnitude of the pixel grayscale variance; The brightness grayscale image is transformed in the frequency domain along the stripe extension direction to obtain a one-dimensional frequency domain signal, and the stripe characteristic frequency and the sidelobe frequency range of the stripe characteristic frequency are determined by the amplitude spectrum analysis of the one-dimensional frequency domain signal. Based on the fringe characteristic frequency and the sidelobe frequency range, a one-dimensional notch filter is constructed in the frequency domain corresponding to the one-dimensional frequency domain signal, and the one-dimensional notch filter is subjected to inverse Fourier transform to obtain an initial time-domain filter coefficient sequence. The initial time-domain filter coefficient sequence is subjected to threshold filtering and normalization based on the statistical variance of the initial time-domain filter coefficient sequence to obtain a one-dimensional time-domain filter kernel with a length shorter than the original sequence length of the initial time-domain filter coefficient sequence. For each row or column of pixels in the brightness grayscale image perpendicular to the stripe direction, a convolution operation is performed sequentially with the one-dimensional temporal filtering kernel to obtain the filtered pixel sequence, thereby obtaining the processed image with stripe noise removed.

2. The method for removing image stripe noise according to claim 1, characterized in that, The process of acquiring raw image data and converting it into a brightness-grayscale image includes: Determine the format of the acquired raw image data; If it is in Bayer format, the complete red, green and blue three-channel color information of each pixel position is first restored by the de-mosaic interpolation algorithm to obtain the RGB image; If it is in RGB format, then directly obtain the RGB image; The luminance value of each pixel in the RGB image is calculated using the weighted formula Gray=0.299×R+0.587×G+0.114×B, and the RGB image is converted into the luminance grayscale image.

3. The method for removing image stripe noise according to claim 1, characterized in that, The step of calculating the pixel grayscale variance in the horizontal and vertical directions of the brightness grayscale image, and determining the stripe direction based on the magnitude of the pixel grayscale variance, includes: The grayscale variance of each row of pixels in the brightness grayscale image is calculated to obtain the horizontal variance sequence, and the grayscale variance of each column of pixels is calculated to obtain the vertical variance sequence. Calculate the average value of the horizontal variance sequence as the overall horizontal variance, and calculate the average value of the vertical variance sequence as the overall vertical variance. Compare the overall variance in the horizontal direction with the overall variance in the vertical direction; If the overall variance in the vertical direction is greater than the overall variance in the horizontal direction and the difference exceeds a preset judgment threshold, then the stripe direction is determined to be horizontal. If the overall variance in the horizontal direction is greater than the overall variance in the vertical direction and the difference exceeds the judgment threshold, then the stripe direction is determined to be vertical.

4. The method for removing image stripe noise according to claim 1, characterized in that, The step of performing a frequency domain transformation on the brightness grayscale image along the extension direction of the stripes to obtain a one-dimensional frequency domain signal includes: When the direction of the stripes is determined to be horizontal: A two-dimensional Fourier transform is performed on the brightness grayscale image to obtain a two-dimensional frequency domain image; the data of each row of the two-dimensional frequency domain image are accumulated or averaged to obtain the one-dimensional frequency domain signal. Alternatively, the pixels of each row of the brightness grayscale image are accumulated to obtain a one-dimensional spatial sequence; a one-dimensional Fourier transform is performed on the one-dimensional spatial sequence to obtain the one-dimensional frequency domain signal. When the direction of the stripes is determined to be vertical: A two-dimensional Fourier transform is performed on the brightness grayscale image to obtain a two-dimensional frequency domain image; the data in each column of the two-dimensional frequency domain image are accumulated or averaged to obtain the one-dimensional frequency domain signal. Alternatively, the pixels in each column of the brightness grayscale image are accumulated to obtain a one-dimensional spatial sequence; a one-dimensional Fourier transform is performed on the one-dimensional spatial sequence to obtain the one-dimensional frequency domain signal.

5. The method for removing image stripe noise according to claim 1, characterized in that, The step of determining the fringe characteristic frequency and the sidelobe frequency range of the fringe characteristic frequency through amplitude spectrum analysis of the one-dimensional frequency domain signal includes: Calculate the amplitude spectrum of the one-dimensional frequency domain signal; Set the DC component corresponding to the zero-frequency position in the amplitude spectrum to zero; Find the peak with the largest amplitude in the amplitude spectrum after setting it to zero, and extract the frequency value corresponding to the peak as the fringe feature frequency. The sidelobe width is determined based on the half-width at half maximum (FWHM) of the spectral peak with the maximum amplitude, and the sidelobe frequency range is defined as: [fringe characteristic frequency − sidelobe width, fringe characteristic frequency + sidelobe width], with the fringe characteristic frequency as the center frequency.

6. The method for removing image stripe noise according to claim 1, characterized in that, The step of constructing a one-dimensional notch filter in the frequency domain corresponding to the one-dimensional frequency domain signal based on the fringe characteristic frequency and the sidelobe frequency range includes: The one-dimensional notch filter is constructed based on the Gaussian function, and its frequency domain expression is as follows: H(f)=1−exp(−(f−f0)² / (2σ²)); Where f0 is the fringe feature frequency, and σ is the Gaussian kernel width; The Gaussian kernel width σ is adaptively calculated based on the sidelobe width, and the calculation formula is as follows: σ = Δf / 1.177, where Δf is the sidelobe width; This makes the gain of the one-dimensional notch filter approach 0 in the sidelobe frequency range and approach 1 outside the sidelobe frequency range.

7. The method for removing image stripe noise according to claim 1, characterized in that, The step of performing threshold filtering and normalization on the initial time-domain filter coefficient sequence based on the statistical variance of the initial time-domain filter coefficient sequence to obtain a one-dimensional time-domain filter kernel with a length shorter than the original sequence length of the initial time-domain filter coefficient sequence includes: Calculate the overall statistical variance of the initial time-domain filter coefficient sequence; The screening threshold is dynamically determined based on the overall statistical variance, wherein the larger the overall statistical variance, the larger the screening threshold is set to. Starting from the center of the initial time-domain filter coefficient sequence, the system iterates towards both ends to select coefficients whose absolute value is greater than the selection threshold, thus forming an effective coefficient sequence. The effective coefficient sequence is normalized so that the sum of all coefficients is 1, and the normalized effective coefficient sequence is used as the one-dimensional time-domain filter kernel.

8. The method for removing image stripe noise according to claim 7, characterized in that, The process of traversing and filtering from the center of the initial time-domain filter coefficient sequence to both ends also includes at least one of the following termination conditions: When the absolute value of a preset number of coefficients does not exceed the filtering threshold for the first time in any one end, further traversal to that end is stopped, and all subsequent coefficients of that end are discarded. Alternatively, when the number of selected coefficients reaches the preset maximum filter kernel length, the filtering stops, and the currently selected coefficients form the effective coefficient sequence.

9. The method for removing image stripe noise according to claim 1, characterized in that, The step of sequentially convolving each row or column of pixel sequences in the brightness-grayscale image perpendicular to the stripe direction with the one-dimensional temporal filter kernel includes: When the stripe direction is determined to be horizontal, the convolution operation is performed on each column of pixel sequence of the brightness grayscale image with the one-dimensional temporal filter kernel. When the direction of the stripes is determined to be vertical, the convolution operation is performed on each row of pixel sequence of the brightness grayscale image with the one-dimensional temporal filter kernel.

10. An electronic device, characterized in that, include: The memory stores execution instructions; as well as A processor that executes execution instructions stored in the memory, causing the processor to perform the method of any one of claims 1 to 9.

Images