A method for detecting strip noise of a panchromatic image of a high-resolution remote sensing satellite

Abstract

The application provides a panchromatic image strip noise detection method of a high-resolution remote sensing satellite. The method calculates and analyzes indexes such as the discrete coefficient and the kurtosis coefficient of the overall and local gradients of the image, and utilizes threshold segmentation technology to detect and locate the position where the strip noise exists. The method is simple and practical, has strong universality, and can quickly and accurately identify the strip noise in the image.

Description

Technical Field

[0001] This invention belongs to the field of satellite remote sensing technology, and in particular relates to a method for detecting stripe noise in panchromatic images of high-resolution remote sensing satellites. Background Technology

[0002] The Gaofen series satellites, with their advantages of high viewing angle, wide observation range, strong survivability, and long-term stable operation, are now widely used in various sectors, including land use surveys, mineral resource development status surveys and monitoring, environmental atmospheric and water environment monitoring, agricultural crop yield estimation and growth monitoring, water conservancy and flood disaster monitoring and water conservancy facility monitoring, and earthquake disaster monitoring, possessing significant economic and military value. High resolution includes high spatial resolution, high temporal resolution, high spectral resolution, and high radiometric resolution (generally, high resolution mainly refers to high spatial resolution). However, as satellites age, remote sensing images are highly susceptible to interference and influence from various noise sources during generation and transmission, leading to a deterioration in image quality. Stripe noise is a common noise phenomenon in Gaofen series satellite remote sensing images, manifested as significant color differences or grayscale variations between adjacent columns or rows. This is caused by various factors, including differences in the optical properties of the sensor's charge-coupled device (CCD) elements, scanning mechanical motion, and sensor platform characteristics. Differences in detector and amplification gain, along with detector non-uniformity, lead to inconsistent responses to the same input radiation energy. This results in uneven grayscale in the pushbroom-generated remote sensing images, causing stripe noise in the images. Furthermore, environmental factors during satellite operation cause thermal disturbances to the arrangement of optical components, resulting in slight differences that also manifest as stripe noise. Stripe noise severely hinders the application of high-resolution satellite remote sensing imagery across various industries, particularly in image interpretation and information extraction, preventing it from fulfilling its intended purpose and significantly impacting subsequent image processing. Therefore, there is an urgent need to research a method for detecting stripe noise in high-resolution satellite imagery, enabling its practical application and large-scale, engineered stripe noise removal.

[0003] Over the past few decades, many scholars at home and abroad have conducted research on strip noise. These methods mainly fall into two categories: (1) One category is relative radiometric correction methods based on the statistical characteristics of the image itself, represented by histogram matching and moment matching methods, which are normalization and matching methods for image gray value features. This type of method is based on the assumption of statistical consistency. When the number of images obtained is small, it has no statistical significance and the denoising effect is poor. This type of method is generally easy to implement, but requires that the ground objects in the noisy image are evenly distributed. (2) The other category is filtering methods, including spatial domain filtering and frequency domain filtering. Spatial domain filtering requires designing a suitable spatial filter and directly convolving it with the image in the spatial domain, but this will cause image blurring, decreased edge contrast, etc.; frequency domain filtering treats the vertical stripes as high-frequency periodic noise. It is necessary to first perform Fourier transform or wavelet transform on the image to separate the noise components and then inverse transform to the spatial domain to finally obtain the denoised image. The main problem with this type of method is that it is easy to remove useful information in non-strip frequencies, which easily loses image details, that is, it has a negative effect on non-strip noise areas. Moreover, most existing literature studies strip noise with specific stripe characteristics, such as linear noise stripes being only one or two pixels wide, or periodically arranged in a certain pattern on the image, making their location relatively easy to determine. However, most challenging strip noises appear without any discernible pattern in their location, and their stripe width is uncertain, ranging from a few pixels to hundreds or thousands of pixels. Therefore, this invention proposes a strip noise detection method based on high-resolution remote sensing satellites, targeting strip noise with different types and characteristics. Summary of the Invention

[0004] The purpose of this invention is to address the problems in existing technologies by proposing a method for detecting stripe noise in panchromatic images from high-resolution remote sensing satellites. This method calculates and analyzes indicators such as the dispersion coefficient and kurtosis coefficient of the overall and local gradients of the image, and utilizes threshold segmentation techniques to detect and locate the position of stripe noise. It is simple, practical, and highly universal, enabling rapid and accurate identification of stripe noise in images.

[0005] This invention is achieved through the following technical solution: This invention proposes a method for detecting stripe noise in panchromatic images from high-resolution remote sensing satellites, the method comprising the following steps:

[0006] Step 1: Detect the overall grayscale distribution of the image, and observe the changes in the overall grayscale gradient and positive and negative gradient values ​​from the perspectives of the discrete coefficient and the kurtosis coefficient.

[0007] Step 2: Detect whether the maximum gradient value in a local region of the image shows a concentrated trend. In this step, if certain rules are met, i.e. the count exceeds half and the gradient weighting value exceeds the threshold, it indicates that there is a significant concentrated gray-level change trend in that column of the image, i.e., the image has strip noise. Based on the count and the result of the gradient weighting value, the degree of noise impact on the image is classified as follows: A is the case of no strip noise, B is the case of possible slight strip noise, which requires further manual confirmation, and C is the case of obvious and non-negligible strip noise.

[0008] Furthermore, in step one, the image is preprocessed, including smooth transition processing, fusion processing, and histogram equalization stretching processing.

[0009] Furthermore, in step one, it is assumed that the total number of pixels in the image is N, the total number of gray levels is L, and the value of the k-th gray level is s. k The image has gray levels s k The number of pixels is n k Then the gray level in the image is s k The probability of a pixel appearing is:

[0010]

[0011] By converting the probability density function into a frequency function, the transformation function of the image is derived as follows:

[0012]

[0013] Finally, the mapping relationship of the new grayscale value is determined. k =int[(L-1)*t k ].

[0014] Furthermore, the calculation of the discrete coefficients is specifically as follows: for each pixel in the image, calculate the magnitude of its horizontal gradient and retain its sign to form the gradient matrix Mat. diff Then, the mean of each column in the gradient matrix is ​​calculated and its absolute value is taken to obtain the column gradient mean vector Vec. grad Therefore, based on the column gradient mean vector Vec grad The discrete coefficient values ​​are obtained.

[0015] Furthermore, by calculating Vec grad Discrete coefficient values To determine whether the gradient change of the image is continuous and uniform; if V > 1.3, then tag = C, that is, the noise level is C, there is obvious strip noise, and the detection process ends; if 1.3 ≥ V > 1.1, tag = B, and proceed to the step of calculating the kurtosis coefficient; if V ≤ 1.1, then proceed to the step of calculating the kurtosis coefficient.

[0016] Furthermore, the calculation of the kurtosis coefficients is specifically as follows: For the obtained gradient matrix Mat... diff Only keep the values ​​with positive gradients and set the rest to 0, thus obtaining the column gradient mean vector Vec. grad Therefore, based on the column gradient mean vector Vec grad The kurtosis coefficient value is obtained.

[0017] Furthermore, by calculating the column gradient mean vector Vec grad kurtosis coefficient The gradient peak value is used to measure the level of the gradient peak. If P > 8, then tag = C and the detection process ends. If 5 < P ≤ 8, tag = B and proceed to step two. If P ≤ 5, proceed to step two.

[0018] Furthermore, in step two, the image is uniformly divided into m*n = s blocks, resulting in n columns. Each column contains m blocks whose x-coordinates lie in the same interval. For each block, the column gradient mean vector Vec of the positive and negative gradients is calculated separately. grad , for Vec grad Sort the data from largest to smallest, and record the column indices of the top k columns into a Set. topk Query the m region records in each column of the Set. topk For columns that appear more than g times, iterate through and check them: iterate through the n blocks containing the column and calculate the mean gradient vector Vec of the columns in each block. grad The gradient mean of the column containing the index column. col ,calculate Calculate COEF block =P block *grad col Record the COEF block The column number of the maximum value (position) col and the corresponding block information, position col That is, the column number where the strip noise is most obvious; if rule one is satisfied: P block >1.6 and coef block If the value is greater than 400, then iterate through position. col For all n blocks in this column, calculate the coef value. block >170 or P block count of blocks greater than 2 block If count block ≥g, and also satisfies rule two: coef block >2300 or P block >9 or 2400-count block*100 < coef block If the count exceeds g but does not meet rule two, mark tag = B; if no block count is entered or the count is less than g and tag = B was not marked before, mark tag = A; otherwise, retain tag = B.

[0019] This invention proposes an electronic device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the method for detecting strip noise in panchromatic images of high-resolution remote sensing satellites.

[0020] This invention proposes a computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the steps of the method for detecting stripe noise in panchromatic images of a high-resolution remote sensing satellite.

[0021] The beneficial effects of this invention are:

[0022] This invention presents a method for detecting stripe noise in panchromatic images from high-resolution remote sensing satellites. Compared to traditional detection schemes based on histogram matching or deep learning, this method is not limited by the distribution of ground features or the uniformity of the imaging system's linear response; it does not compromise image clarity; it can achieve accurate localization regardless of the width or regularity of the stripe noise's distribution; and it has no strict requirements on data samples, low computational complexity, significantly reduces manual screening operations, and minimizes time and resource consumption. In summary, this invention provides a more efficient, accurate, and precise triple-protection detection algorithm with good robustness and universality, and can be directly applied to different satellite image data. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0024] Figure 1 This is a schematic diagram of a single line with noise.

[0025] Figure 2 This is a schematic diagram of narrow strip noise.

[0026] Figure 3 This is a schematic diagram of wide strip noise.

[0027] Figure 4 This is a schematic diagram of strip noise level A.

[0028] Figure 5 This is a schematic diagram of strip noise level B.

[0029] Figure 6 This is a schematic diagram of strip noise level C.

[0030] Figure 7 This is a flowchart of a method for detecting stripe noise in panchromatic images of high-resolution remote sensing satellites, as proposed in this invention. Detailed Implementation

[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] Combination Figures 1-7 This invention proposes a method for detecting stripe noise in panchromatic images from high-resolution remote sensing satellites, the method comprising the following steps:

[0033] Step 1: Detect the overall grayscale distribution of the image, and observe the changes in the overall grayscale gradient and positive and negative gradient values ​​from the perspectives of the discrete coefficient and the kurtosis coefficient.

[0034] Step 2: Detect whether the maximum gradient value in a local region of the image shows a concentrated trend. In this step, if certain rules are met, i.e. the count exceeds half and the gradient weighting value exceeds the threshold, it indicates that there is a significant concentrated gray-level change trend in that column of the image, i.e., the image has strip noise. Based on the count and the result of the gradient weighting value, the degree of noise impact on the image is classified as follows: A is the case of no strip noise, B is the case of possible slight strip noise, which requires further manual confirmation, and C is the case of obvious and non-negligible strip noise.

[0035] In step one, the image is preprocessed, including smoothing transitions, fusion processing, and histogram equalization stretching. A standard panchromatic image produced by a high-resolution remote sensing satellite is a 16-bit depth, single-channel grayscale DN image file. Satellites often use multi-sensor stitching to increase the imaging swath width; under normal circumstances, the overlapping areas of multiple sensors have been smoothed, and the stitching lines have been eliminated through fusion.

[0036] Due to the influence of geographical features and seasonal weather, the grayscale distribution of remote sensing satellite images often exhibits a concentrated trend. To unify the measurement standards of all images to be inspected, histogram equalization stretching is required for each image to achieve a balanced grayscale distribution and standardized dimensions. Histogram equalization is a simple and effective image enhancement technique that alters the grayscale of each pixel in an image by changing its histogram. It is primarily used to enhance the contrast of images with a small dynamic range. By employing histogram equalization, the dynamic range of grayscale value differences between pixels can be increased, enhancing the boundary definition of bands and thus highlighting the difference between noisy and normal areas, emphasizing the location of noise.

[0037] Assume the image has N pixels in total, L gray levels in total, and the value of the k-th gray level is s. k The image has gray levels s k The number of pixels is n k Then the gray level in the image is s k The probability (or frequency) of a pixel appearing is:

[0038]

[0039] By converting the probability density function into a frequency function, the transformation function of the image is derived as follows:

[0040]

[0041] Finally, the mapping relationship of the new grayscale value is determined. k =int[(L-1)*t k ].

[0042] The calculation of the discrete coefficients is as follows: for each pixel in the image, calculate the magnitude of its horizontal gradient and retain its sign, forming the gradient matrix Mat. diff Then, the mean of each column in the gradient matrix is ​​calculated and its absolute value is taken to obtain the column gradient mean vector Vec. grad Therefore, based on the column gradient mean vector Vec grad The discrete coefficient values ​​are obtained.

[0043] By calculating Vec grad Discrete coefficient values This is used to determine whether the gradient change of an image is continuous and uniform. Generally, in a normal image without stripe noise, the gradient will be concentrated around 0, with small fluctuations. Even if there are peaks, they will be relatively gentle, gradually rising rather than abruptly reaching the peak. Therefore, when the dispersion coefficient exceeds the threshold, it means that the gradient distribution is not gentle, the jitter is severe, the peak value is high, and the image has obvious stripe noise. If V > 1.3, then tag = C, that is, the noise level is C, there is obvious stripe noise, and the detection process ends; if 1.3 ≥ V > 1.1, tag = B is marked, and the process proceeds to the step of calculating the kurtosis coefficient; if V ≤ 1.1, the process proceeds to the step of calculating the kurtosis coefficient.

[0044] The calculation of the kurtosis coefficient is specifically as follows: For the obtained gradient matrix Mat... diff Only keep the values ​​with positive gradients and set the rest to 0, thus obtaining the column gradient mean vector Vec. grad Therefore, based on the column gradient mean vector Vec grad The kurtosis coefficient value is obtained.

[0045] By calculating the column gradient mean vector Vec grad kurtosis coefficient The gradient peak value is used to measure its strength. This invention retains a gradient in a single direction to prevent the averaging process from easily canceling out noise gradient directions due to variations in the brightness of ground features. Furthermore, it notes that the gradient direction at the location of the strip noise remains consistent across the entire image column, providing a theoretical basis for calculating only the effective gradient. Similarly, this invention calculates the kurtosis coefficient of the column gradient mean vector when the gradient is negative. If the kurtosis coefficient exceeds a threshold, it indicates that the peak value deviates significantly from the center, the image's grayscale change occurs with a significant step in the current column, and the image exhibits obvious strip noise. If P > 8, tag = C, and the detection process ends; if 5 < P ≤ 8, tag = B, and proceed to step two; if P ≤ 5, proceed to step two.

[0046] In step two, the image is uniformly divided into m*n = s blocks, resulting in n columns. Each column contains m blocks whose x-coordinates lie in the same interval. For each block, the column gradient mean vector Vec of the positive and negative gradients is calculated separately. grad , for Vec grad Sort the data from largest to smallest, and record the column indices of the top k columns into a Set. topk Query the m region records in each column of the Set. topk For columns that appear more than g times, iterate through and check them: iterate through the n blocks containing the column and calculate the mean gradient vector Vec of the columns in each block. grad The gradient mean of the column containing the index column. col ,calculate Calculate COEF block =P block *grad col Record the COEF block The column number of the maximum value (position) col and the corresponding block information, position col This refers to the column number where the strip noise is most prominent, which is generally the boundary line of the strip noise; if rule one is satisfied: P block >1.6 and coef block If the value is greater than 400, then iterate through position. col For all n blocks in this column, calculate the coef value. block >170 or P block count of blocks greater than 2 block If count block ≥g, and also satisfies rule two: coef block >2300 or P block >9 or 2400-count block *100 < coef block， If tag = C, the detection process ends; if the count exceeds g but does not meet rule 2, tag = B; if no block count is entered or the count is less than g and tag = B was not marked before, tag = A is marked, otherwise tag = B is retained.

[0047] The method described in this invention first detects the overall grayscale distribution of the image, observing the changes in the overall grayscale gradient and positive and negative gradient values ​​from the perspectives of the discrete coefficient and kurtosis coefficient. However, some stripe noise is only obvious in local locations and may not appear in the entire column of the image, or it may be difficult to distinguish with the naked eye in other locations. Using the average value to replace the overall trend actually dilutes the presence of stripe noise. Therefore, the second step is to detect whether the maximum gradient value in a local area of ​​the image shows a concentrated trend. In this step, if certain rules are met, such as the count exceeding half and the gradient weighting value exceeding a threshold, it indicates that there is a significant concentrated grayscale change trend in that column of the image, that is, the image contains stripe noise. Based on the count and the gradient weighting value, the degree of noise impact on the image is classified: A is the case of no stripe noise, B is the case of possible slight stripe noise that requires further manual confirmation, and C is the case of obvious and non-negligible stripe noise. This step not only has a significant detection effect on local stripe noise, but also avoids the influence of grayscale jump changes in local objects. In statistics, the average value is greatly affected by extreme points. In images, this is reflected in the edges of coastlines, streets, or clouds, where the grayscale values ​​change significantly, which can affect the gradient mean of the entire column. Therefore, this block-based strategy can reduce the influence range of abnormal deviations to a sufficiently small extent while dividing the detection range into sufficiently fine sections. It can be said to be the most powerful guarantee in this detection scheme.

[0048] The effectiveness of the method described in this invention is illustrated below using a test sample dataset. The test sample dataset used was the Jilin-1 Gaofen 03D series satellite for stripe noise detection. It contained 735 normal images without stripe noise and 760 images with stripe noise. Each image was approximately 24000×24000 pixels. The optimal parameters were determined experimentally to be m=30, n=40, k=10, and g=14. The test results showed that among the normal images without stripe noise, 2 were misclassified as Grade C, 57 as Grade B, and 676 were correctly classified as Grade A. Among the images with stripe noise, 42 were classified as Grade A. Since Grade B images would undergo further manual verification to avoid data loss, only images with stripe noise were classified as Grade A (false negatives), and images without stripe noise were classified as Grade C (false positives). The test results showed a false negative rate of 2.81%, a false positive rate of 0.13%, and a classification accuracy of 97.06%. Absolute correct classification means classifying images without striped noise into level A, and images with striped noise into level B or C. Therefore, the absolute classification accuracy was 91.97% on the dataset without striped noise, 94.47% on the dataset with striped noise, and 93.24% on the entire dataset.

[0049] This invention proposes an electronic device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the method for detecting strip noise in panchromatic images of high-resolution remote sensing satellites.

[0050] This invention proposes a computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the steps of the method for detecting stripe noise in panchromatic images of a high-resolution remote sensing satellite.

[0051] The memory in this application embodiment can be volatile memory or non-volatile memory, or it can include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (SLDRAM), and direct rambus RAM (DRRAM). It should be noted that the memory used in the methods described in this invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0052] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., high-density digital video discs (DVDs)), or semiconductor media (e.g., solid-state disks (SSDs)).

[0053] In implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software. The steps of the method disclosed in the embodiments of this application can be directly implemented by a hardware processor, or by a combination of hardware and software modules in the processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, detailed descriptions are omitted here.

[0054] It should be noted that the processor in the embodiments of this application can be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method embodiments can be completed by the integrated logic circuitry in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied as being executed by a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above methods.

[0055] The above provides a detailed description of a method for detecting stripe noise in panchromatic images of high-resolution remote sensing satellites proposed in this invention. Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method for detecting strip noise in a panchromatic image of a high-resolution remote sensing satellite, characterized in that, The method comprises the following steps: Step one, detecting the gray distribution state of the whole image, observing the whole gray gradient and the change of positive and negative gradient values from two angles of discrete coefficient and kurtosis coefficient respectively; Step two, detecting whether the local area of the image has a centralized trend of gradient maximum value, if certain rules are met, it is indicated that there is an obvious centralized gray change trend in a column of the image, and the image has band noise; the degree of influence of the image by noise is classified, specifically: tag=A is the case without band noise, tag=B is the case that there may be slight band noise, which needs to be further confirmed by human, and tag=C is the case of obvious band noise that cannot be ignored; In step two, the image is uniformly divided into m*n=s blocks, resulting in n columns. Each column contains m blocks whose x-coordinates lie in the same interval. The column gradient mean vectors for both positive and negative gradients are calculated separately for each block. ,right Sort by largest to smallest, and record the top-ranked items. Column number to Query the m region records in each column. The frequency of occurrence in the middle exceeds Column number of the next Perform a traversal check: Traversal Where block Calculate the column gradient mean vector for each current block. ,index Gradient mean of the column ,calculate ,calculate Record The column number containing the maximum value and the corresponding block information, The column number where the stripe noise is most pronounced; if rule one is satisfied: >1.6 and If the value is greater than 400, then iterate through the list. All of this column Block, statistics of >170 or Number of blocks >2 ,like It also satisfies rule two: >2300 or >9 or If tag=C, the detection process ends; if Exceed However, if rule two is not met, tag=B is set; if it is not included in the block count or Less than If tag=B was not previously marked, then tag=A; otherwise, tag=B will be retained.

2. The method of claim 1, wherein, In step one, the image is preprocessed, including smooth transition processing, fusion processing and histogram equalization stretching processing.

3. The method of claim 2, wherein, In step one, assume that the total pixels in the image are N, the total number of gray levels is L, the value of the kth gray level is , the number of pixels with the gray level in the image is , and the probability of the pixels with the gray level in the image is: The probability density function is converted into the frequency function, and the transformation function of the image is derived as follows: Finally determine the mapping of new gray value correspondence .

4. The method of claim 1, wherein, The calculation of the discrete coefficient is specifically: for each pixel point in the image, the gradient size in the horizontal direction is calculated and the sign is kept to form a gradient matrix Then, the mean value of each column in the gradient matrix is calculated and the absolute value is taken to obtain a column gradient mean value vector According to the column gradient mean value vector The discrete coefficient value is obtained.

5. The method of claim 4, wherein, The gradient change of the image is judged to be continuous and uniform by calculating the discrete coefficient value of If , it is determined that tag=C, the noise level is C, there is obvious band noise, and the detection process ends. If , tag=B is marked, and the peak state coefficient calculation step is entered. If , the peak state coefficient calculation step is entered.​ 6. The method of claim 5, wherein, The calculation of the kurtosis coefficient is specifically as follows: the obtained gradient matrix is only the values of the positive gradient are retained, and the rest are set to 0, and a column gradient mean vector is also obtained , so that the kurtosis coefficient value is obtained according to the column gradient mean vector .

7. The method of claim 6, wherein, By calculating the mean vector of column gradients The peakiness coefficient of the peakiness coefficient of the gradients; if P>8, then tag=C, and the detection process ends. If 5 8.An electronic device comprising a memory and a processor, the memory storing a computer program, wherein, The processor executes the computer program to realize the steps of the method in any one of claims 1-7.

9. A computer readable storage medium for storing computer instructions, characterized in that, The computer instructions are executed by the processor to realize the steps of the method in any one of claims 1-7.

Images