Method for removing stripe noise of satellite image

Abstract

The invention relates to the technical field of earth science, in particular to a method for removing stripe noise of a satellite image, which comprises the following steps of: acquiring the satellite image, and separating the satellite image into a high-frequency image and a low-frequency image by using wavelet transform; fitting analysis is carried out on a non-noise area corresponding to the surface feature area and the panchromatic image, a ratio relation between the non-noise area and a panchromatic image DN value is established based on a fitting result, the noise area is corrected based on the ratio relation and the panchromatic image, and a corrected image corresponding to the surface feature area is obtained; and combining the corrected image with the denoised high-frequency image to carry out wavelet inverse transformation to obtain a denoised satellite image, so that the stripe noise of the blue-light wave band of the satellite image can be removed, the influence of the complex stripe noise on the satellite image is reduced, and the ground feature information is reserved to the greatest extent.

Description

Technical Field

[0001] The present application relates to the field of earth science technology, and in particular to a method for removing stripe noise from satellite images. Background Art

[0002] Small satellites have the advantages of low cost, short development cycle, and high flexibility. They can respond to needs quickly and realize customized tasks. At the same time, small satellites can also realize multi-satellite networking, improve coverage and data collection efficiency, so small satellites have broad application prospects in the fields of earth observation, communication navigation, scientific research, etc. However, due to the limitations of factors such as cost and technology, in terms of geometric accuracy correction and radiation quantitative standard processing, domestic small satellites are still a certain distance away from the maturity of European and American small satellites, and there are obvious defects. The visible light sensor carried by the Ice Road Satellite BNU-1 has the advantages of high coverage and high spatial resolution, and can obtain high-quality optical remote sensing images. However, due to the distribution of environmental factors such as darkness, noise, and heat, the radiation response of CMOS sensors is also unbalanced, which often leads to nonlinear radiation errors in remote sensing images. Therefore, before applying visible light remote sensing images, it is necessary to evaluate and correct the radiation measurement errors. Radiation correction is an important prerequisite for ensuring the quality and accurate application of satellite images, and the presence of noise will seriously affect the analysis and recognition of satellite images. In order to reduce the influence of stripe noise, denoising methods can be adopted to improve image quality. The most commonly used method is to use filters to post-process satellite images. The removal method of stripe noise can be discussed from two perspectives: statistics and filtering. In statistical methods, histogram matching and moment matching are two commonly used methods. They remove stripe noise by adjusting the histogram distribution or moment characteristics of the image, while in filtering methods, wavelet transform and spatial domain-frequency domain conversion methods perform well. By processing stripe noise, we can better retain the detailed information of the image, making it clearer and more accurate.

[0003] However, in the prior art, even the corrected image still has stripe noise in the blue light band, which will affect the clarity and quality of the image and bring difficulties to image processing and analysis.

[0004] Therefore, how to better remove the stripe noise in the blue light band of satellite images is an urgent problem to be solved. Summary of the invention

[0005] The present application provides a method for removing stripe noise from satellite images, which can remove the stripe noise in the blue light band of satellite images based on wavelet transform, reduce the impact of complex stripe noise on satellite images, and retain the ground object information to the maximum extent. The method also has good universality.

[0006] In order to achieve the above object, the present invention adopts the following technical scheme:

[0007] In a first aspect, the present invention provides a method for removing stripe noise from satellite images, the method comprising:

[0008] Collect satellite images, and separate the satellite images into high-frequency images and low-frequency images using wavelet transform;

[0009] Performing denoising on the high-frequency image to obtain a denoised high-frequency image;

[0010] Splitting the low-frequency image to obtain a plurality of ground feature regions;

[0011] Extracting strip noise from the feature area, obtaining image noise extraction results corresponding to each direction of the feature area, and obtaining a non-noise area and a noise area corresponding to the feature area;

[0012] Performing fitting analysis on the non-noise area corresponding to the feature area and the panchromatic image, establishing a ratio relationship between the DN value of the non-noise area and the panchromatic image based on the fitting result, and correcting the noise area based on the ratio relationship and the panchromatic image to obtain a corrected image corresponding to the feature area;

[0013] The corrected image is combined with the denoised high-frequency image to perform inverse wavelet transform to obtain a denoised satellite image.

[0014] In a preferred example of the present application, it can be further configured that the use of wavelet transform to separate the satellite image into high-frequency image and low-frequency image includes:

[0015] The satellite image is separated into a low-frequency image and a high-frequency image using a Haar wavelet function, wherein the high-frequency image includes a high-frequency part image in the horizontal direction, a high-frequency part image in the vertical direction, and a high-frequency part image in the diagonal direction.

[0016] In a preferred example of the present application, it can be further configured that the denoising process is performed on the high-frequency image to obtain the denoised high-frequency image, including:

[0017] The soft threshold method is used to select the threshold, and the high-frequency image is subjected to threshold processing to obtain the denoised high-frequency image. The calculation formula of the soft threshold is:

[0018]

[0019] Among them, w is the wavelet coefficient, T is the threshold, which is 0.5 times the image standard deviation, m is the coefficient mean, and n is the number of calculations.

[0020] In a preferred example of the present application, it can be further configured that the low-frequency image is split to obtain multiple ground feature areas, including:

[0021] Based on the DN value of the satellite image, the low-frequency image is split into regions using a natural breakpoint method to obtain a plurality of land feature regions.

[0022] In a preferred example of the present application, it can be further configured that the strip noise of the ground feature area is extracted, including

[0023] The Sobel operator is used to calculate the horizontal and vertical gradient values ​​of the object region, and the strip noise of the object region is extracted based on the gradient values.

[0024] In a preferred example of the present application, it can be further configured that the Sobel operator is used to calculate the horizontal and vertical gradient values ​​of the feature area, and the strip noise of the feature area is extracted based on the gradient values, including:

[0025] Calculate the gradient value in the horizontal direction of the feature area, obtain the area where the change of the gradient amplitude is greater than a preset threshold, and use the area as the approximate strip noise;

[0026] The edge of the object is obtained based on the gradient value in the vertical direction, and the edge of the object in the approximate strip noise is removed to obtain the strip noise of the object area.

[0027] In a second aspect, the present application provides a device for removing stripe noise from satellite images, the device comprising:

[0028] A data acquisition module, used for acquiring satellite images, and separating the satellite images into high-frequency images and low-frequency images using wavelet transform;

[0029] A high-frequency denoising module, used to perform denoising on the high-frequency image to obtain a denoised high-frequency image;

[0030] A low-frequency denoising module is used to split the low-frequency image to obtain multiple ground feature areas;

[0031] Extract the strip noise of the feature area, obtain the image noise extraction results corresponding to each direction of the feature area, and obtain the non-noise area and the noise area corresponding to the feature area; perform fitting analysis on the non-noise area corresponding to the feature area and the panchromatic image, establish a ratio relationship between the DN value of the non-noise area and the panchromatic image based on the fitting result, correct the noise area based on the ratio relationship and the panchromatic image, and obtain a corrected image corresponding to the feature area;

[0032] The output module is used to perform inverse wavelet transform on the corrected image in combination with the denoised high-frequency image to obtain a denoised satellite image.

[0033] In a third aspect, the present application provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the method for removing stripe noise from satellite images as described in any one of the above items are implemented.

[0034] In a fourth aspect, the present application provides a computer-readable storage medium having a program stored thereon, wherein when the program is executed by a processor, the method for removing stripe noise from satellite images as described in any one of the above items is implemented.

[0035] In a fifth aspect, the present application provides a computer program product, comprising computer instructions, which, when executed by a processor, implement the steps of the method for removing stripe noise from satellite images as described in any one of the above items.

[0036] In summary, compared with the prior art, the technical solution provided in the embodiment of the present application has at least the following beneficial effects:

[0037] The present application provides a method for removing stripe noise from satellite images. The method comprises the following steps: collecting satellite images, separating the satellite images into high-frequency images and low-frequency images by using wavelet transform; performing denoising on the high-frequency images to obtain denoised high-frequency images; splitting the low-frequency images to obtain a plurality of object regions; extracting stripe noise from the object regions to obtain image noise extraction results corresponding to each direction of the object regions, and obtaining non-noise regions and noise regions corresponding to the object regions; performing fitting analysis on the non-noise regions corresponding to the object regions and the panchromatic images, establishing a ratio relationship between the DN values ​​of the non-noise regions and the panchromatic images based on the fitting results, and correcting the noise regions based on the ratio relationship and the panchromatic images to obtain a corrected image corresponding to the object regions; performing inverse wavelet transform on the corrected images in combination with the denoised high-frequency images to obtain denoised satellite images. The method can remove stripe noise in the blue light band of satellite images, reduce the influence of complex stripe noise on satellite images, and retain object information to the maximum extent. The method also has good universality. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A flowchart of a method for removing stripe noise from satellite images provided in one embodiment of the present application.

[0039] Figure 2 A schematic diagram of a high- and low-frequency signal separation image of a method for removing strip noise from a satellite image provided by an embodiment of the present application.

[0040] Figure 3 A schematic diagram of high-frequency processing results of a method for removing stripe noise from satellite images provided in one embodiment of the present application.

[0041] Figure 4 A schematic diagram of low-frequency image region decomposition of a satellite image strip noise removal method provided in one embodiment of the present application.

[0042] Figure 5 A schematic diagram of image stripe extraction results of a method for removing stripe noise from satellite images provided in one embodiment of the present application.

[0043] Figure 6 A comparison diagram and a linear relationship diagram of a noisy image and a denoised image of a method for removing strip noise from a satellite image provided in one embodiment of the present application.

[0044] Figure 7 A diagram showing the fitting results of a non-noise area image and a full-color image of a satellite image strip noise removal method provided in one embodiment of the present application.

[0045] Figure 8 A comparison diagram of BNU-1 blue band images before and after denoising, which is a method for removing strip noise from satellite images provided in one embodiment of the present application.

[0046] Fig. 9 A structural diagram of a device for removing stripe noise from satellite images provided in one embodiment of the present application. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0048] In one embodiment of the present application, a method for removing stripe noise from satellite images is provided. Figure 1 As shown, the method includes:

[0049] S100: collecting satellite images, and using wavelet transform to separate the satellite images into high-frequency images and low-frequency images;

[0050] Specifically, the satellite image is a domestic polar small satellite image. First, the high-frequency and low-frequency noise of the satellite image are distinguished by wavelet transform. The specific method is described as follows:

[0051] The formula for wavelet transform is:

[0052]

[0053] in, is a wavelet function, this time the Haar wavelet is used, a and b are scale and translation parameters, and x(τ) is the original signal.

[0054] The formula of Haar wavelet function is as follows:

[0055]

[0056] Where x represents the position in time or space, and ψ(x) is the value of the Haar wavelet function at position x.

[0057] The wavelet transform method is used to divide the image into four parts, namely, the low-frequency signal image, the horizontal high-frequency part image, the vertical high-frequency part image, and the diagonal high-frequency part image.

[0058] After analysis and differentiation, there are obvious stripes in the high-frequency horizontal direction, high-frequency vertical direction or low-frequency image. These stripes are manifestations of noise. The wavelet transform method used in the past mainly processes high-frequency narrow-band noise, and low-frequency wide-band noise cannot be removed.

[0059] S200: performing denoising processing on the high-frequency image to obtain a denoised high-frequency image;

[0060] Specifically, in the wavelet transform, among the four components obtained after processing, part of the noise of the high-frequency image can be directly removed using the wavelet transformation method. Among the four components obtained after the wavelet transform, the general wavelet denoising method is directly used for the high-frequency component (high-frequency image). After the wavelet transform, the signal is usually concentrated on fewer high-frequency wavelet coefficients, while the noise is distributed on the wavelet coefficients of all scales. Therefore, by thresholding the high-frequency wavelet coefficients, the noise can be selectively removed while retaining the signal.

[0061] In addition, for many signals, their wavelet transformed coefficients have a sparse property, that is, most of the wavelet coefficients are very close to zero. This is because the signal usually has only a few significant frequency components in the wavelet domain, while noise will cause the wavelet coefficients of all scales to have large changes. Therefore, by thresholding the wavelet coefficients, the noise coefficient can be set to zero, thereby achieving the effect of denoising.

[0062] This embodiment uses the soft threshold method to select the threshold of low-pass filtering. This method is simple to calculate, easy to implement, and performs well in many cases, so it is widely used in the field of wavelet denoising. The calculation formula of the soft threshold is as follows:

[0063]

[0064] Among them, w is the wavelet coefficient, T is the threshold, which is set to 0.5 times the image standard deviation in this study, m is the coefficient mean, and n is the number of calculations.

[0065] S300: splitting the low-frequency image to obtain multiple ground feature regions;

[0066] Specifically, through data analysis, we found that in low-frequency images, strip noise is very complex, and the degree of strip noise in different areas is different. Even the interference degree of noise on the same strip will vary. Therefore, to remove low-frequency strip noise, it is necessary to first split the image according to the noise characteristics. This study splits the region according to the DN value of the image. The splitting method uses the natural breakpoint method, which can ensure that the DN value of the data within each region is close, and the data between different regions is more distinguishable. The formula is as follows:

[0067] Given a wavelet coefficient array c, whose length is N, assume that c = [c1, c2, ..., c N ], then the threshold selection formula of the natural breakpoint method is as follows:

[0068]

[0069] Where σ is the standard deviation of the wavelet coefficients, The part represents the adjustment of the threshold by the number of wavelet coefficients.

[0070] Finally, the land feature areas corresponding to different landforms are obtained, and the landforms may include ice caps, rocks, or snow.

[0071] S400: extracting strip noise from the feature area, obtaining image noise extraction results corresponding to each direction of the feature area, and obtaining a non-noise area and a noise area corresponding to the feature area;

[0072] In terms of identifying objects and noise, commonly used methods include texture feature recognition and edge extraction. This embodiment uses the Sobel operator for gradient calculation. The Sobel operator is often used in image processing to calculate the gradient of an image in order to detect edges in the image. For a two-dimensional image I, its grayscale value can be expressed as I(x, y), where x and y represent the horizontal and vertical positions in the image, respectively. The Sobel operator is divided into two convolution kernels in the horizontal direction and the vertical direction, respectively denoted as G x and G y :

[0073]

[0074] They are used to calculate the gradient in the horizontal and vertical directions respectively.

[0075] For each pixel I(x,y) in the image, the Sobel operator in the horizontal and vertical directions is convolved with the neighboring pixels around the pixel to obtain the gradient value of the pixel. Formula 8 to Formula 10 are the horizontal gradient calculation formula, the vertical gradient calculation formula and the gradient amplitude that can be used for edge gradient extraction.

[0076]

[0077] According to the gradient value, the image noise extraction results corresponding to the vertical direction and the horizontal direction of the feature area are obtained, and then the non-noise area and the noise area corresponding to the feature area are determined.

[0078] S500: performing fitting analysis on the non-noise area corresponding to the feature area and the panchromatic image, establishing a ratio relationship between the DN values ​​of the non-noise area and the panchromatic image based on the fitting result, and correcting the noise area based on the ratio relationship and the panchromatic image to obtain a corrected image corresponding to the feature area;

[0079] Specifically, the blue band image is used as the noise image processed in this embodiment, and the panchromatic image is used to correct it in combination with the ground feature area separated previously, and the noise area and non-noise area extracted. For the convenience of subsequent description, these data are simply numbered: let the blue band image be B, the panchromatic image be P, the four ground feature areas separated be a, b, c, d, the non-noise area extracted be 1, and the noise area be 2. For example: P a1 It is the intersection range of area a and non-noise area 1 on the full-color image.

[0080] Since the split areas are processed separately, at least two areas are required for each processing to calculate the relationship between points. Here, area a and area b are selected to introduce the process. The specific steps are as follows:

[0081] (1) Use image alignment based on maximizing the enhanced correlation coefficient to coarsely align the images. This step can be modified as needed because the registration method is prone to new deviations in the images due to noise interference.

[0082] (2) In the low-frequency bands of the panchromatic image and the noise image, the intersection of the non-noise area and area a is selected (P a1 With B a1 ) as the reference area, and the intersection of the noise area and area b is taken as the area to be processed (P b2 With B b2 ).

[0083] (3) In P a1 Randomly select n pixels [p1, p2…p n ], calculate its average value and Pb2 A pixel (p m ) and combine it with the linear fitting result between the two images as B a1 The n pixels [b1, b2…b n The average value of ] and B b2 The same pixel (b m ), that is, the ratio relationship that each pixel in the noise image should have in the absence of noise.

[0084] (4) Determine the difference between the relative relationships. If the difference is too large, use the above relationship to correct the noise area, that is, use the ideal ratio relationship to replace the corresponding ratio relationship on the noise image.

[0085] After correction, the corrected image corresponding to the feature area is obtained.

[0086] S600: performing inverse wavelet transform on the corrected image and the denoised high-frequency image to obtain a denoised satellite image.

[0087] Specifically, after the correction is completed, the previously processed high-frequency image (high-frequency component) is subjected to an inverse wavelet transform to obtain the final result, i.e., the denoised satellite image. The inverse wavelet transform formula to be used is:

[0088]

[0089] in C ψ is the constant of the wavelet function, and the other parameters are consistent with Formula 2.

[0090] The noisy image in S100 is replaced with the denoised satellite image, and steps S200-S600 are repeated, and the process is iterated continuously until various indicators of the noisy image tend to be normal or cannot be further improved.

[0091] The number of points n selected in step (3) of step S500 can be adjusted according to the situation. In theory, a larger number is better, because the accuracy of strip extraction cannot be determined, and there may be noise extracted by mistake in the non-strip area. Therefore, this method can reduce the proportion of noise points in random points. The basic number used in this embodiment is 1000, and in some cases it will be adjusted accordingly according to the actual image.

[0092] In this embodiment, the stripe noise in the blue light band of the satellite image can be removed, the influence of the complex stripe noise on the satellite image can be reduced, and the ground object information can be retained to the maximum extent. The method also has good universality.

[0093] In some embodiments, using a Sobel operator to calculate horizontal and vertical gradient values ​​of the feature area, and extracting strip noise of the feature area based on the gradient values, includes:

[0094] Calculate the gradient value in the horizontal direction of the feature area, obtain the area where the change of the gradient amplitude is greater than a preset threshold, and use the area as the approximate strip noise;

[0095] The edge of the object is obtained based on the gradient value in the vertical direction, and the edge of the object in the approximate strip noise is removed to obtain the strip noise of the object area.

[0096] In the specific implementation, for the area with strip noise in the vertical direction, the edge extraction algorithm based on the Sobel operator is adjusted, and the gradient information in the vertical direction is not used for strip extraction, but for mis-extraction elimination. The specific steps are as follows:

[0097] (1) First, calculate the gradient information in the horizontal direction, obtain the areas where the gradient changes greatly, and identify them as approximate stripe noise;

[0098] (2) The general edge of the ground object is obtained through the vertical gradient information, and the edge of the ground object in the strip noise identified in the previous step is eliminated.

[0099] In some embodiments, the accuracy of stripe noise extraction in the vertical direction of the ground feature area is improved by eliminating false extraction.

[0100] In some embodiments, the Greenland satellite image BNU1 taken on August 24, 2020 is taken as an example to illustrate the above technology. The example image is divided into four parts using the wavelet transform method. Figure 2 , Figure 2 (a) is the original BNU-1 image, Figure 2 (b) Figure 2 (c) Figure 2 (d) Figure 2 (e) are the low-frequency signal image, the high-frequency part image in the horizontal direction, the high-frequency part image in the vertical direction, and the high-frequency part image in the diagonal direction after decomposition. It can be seen from the figure that there are obvious stripes in both the high-frequency horizontal direction and the high-frequency vertical direction or the low-frequency image. These stripes are manifestations of noise and are also the high-frequency stripe noise that this embodiment attempts to eliminate.

[0101] Figure 3 It is the result of denoising the high-frequency components. Figure 3 (a) is the horizontal high-frequency image after denoising. Figure 3 (b) is the high-frequency image after vertical denoising. Figure 3(c) is the high-frequency image in the diagonal direction after denoising. It can be seen from the naked eye that the vertical stripes have been significantly reduced, and the denoising result in the high-frequency area is relatively ideal.

[0102] This embodiment uses wavelet transformation to obtain low-frequency images, calculates natural breakpoints, and divides the impact into four areas based on the natural breakpoints: P1-P4, i.e., the ground feature areas. Figure 4 As shown in the figure, the image can be effectively divided into four areas: ice cover, rock, snow, and transition between rock and snow by using gradient changes.

[0103] It can be found that the stripe noise mainly appears in the ice cover area of ​​the image, so this embodiment further removes the image noise based on this.

[0104] Figure 5 The vertical strip extraction results, horizontal strip extraction results and final extraction results of the images selected in this chapter are shown.

[0105] The image after noise removal is compared with the original noise image. Figure 6 , Figure 6 (a) is the original noisy image, Figure 6 (b) is the image after denoising. The image after denoising has almost no obvious noise visible to the naked eye. Figure 6 (a) and Figure 6 (b) The DN value corresponding to the red section line in the two figures, and the curve is drawn Figure 6 (c). Figure 6 In (c), the blue curve is the DN value of the original noise image, and the orange curve is the denoised image. It can be seen that the oscillation amplitude of the blue curve of the original noise image is relatively large, and the standard deviation reaches 62.79, indicating that the noise of this curve is relatively obvious. After denoising, the oscillation amplitude of the orange curve is significantly reduced, showing a gentle trend, and the standard deviation drops to 28.06, a decrease of about 55%.

[0106] The non-noise area of ​​the low-frequency component of the sample image is used for fitting analysis with the panchromatic image. The fitting results are shown in Figure 7 The results show that the coefficient of determination R2 between the two reaches 0.906, indicating that there is a significant linear correlation between the two. Therefore, this fitting result is used to establish the relationship between the non-noise area and the DN value of the full-color image. This relationship is used to correct the image of the de-noised area, and finally the image after the stripe noise is removed is obtained. Figure 8 This shows that the improved wavelet transform method used in this embodiment has a better effect in removing stripe noise.

[0107] This application also provides a stripe noise removal device based on satellite images, please refer to Fig. 9 As shown, the device comprises:

[0108] The data acquisition module 100 is used to acquire satellite images and separate the satellite images into high-frequency images and low-frequency images using wavelet transform;

[0109] A high frequency denoising module 200 is used to perform denoising on the high frequency image to obtain a denoised high frequency image;

[0110] A low-frequency denoising module 300 is used to split the low-frequency image to obtain multiple ground feature regions;

[0111] Extract the strip noise of the feature area, obtain the image noise extraction results corresponding to each direction of the feature area, and obtain the non-noise area and the noise area corresponding to the feature area; perform fitting analysis on the non-noise area corresponding to the feature area and the panchromatic image, establish a ratio relationship between the DN value of the non-noise area and the panchromatic image based on the fitting result, correct the noise area based on the ratio relationship and the panchromatic image, and obtain a corrected image corresponding to the feature area;

[0112] The output module 400 is used to perform inverse wavelet transform on the corrected image and the denoised high-frequency image to obtain a denoised satellite image.

[0113] The functional implementation of each module in the above-mentioned stripe noise removal device based on satellite images corresponds to the steps in the above-mentioned satellite image stripe noise removal method embodiment, and its functions and implementation processes are not repeated here one by one.

[0114] The present application also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the method for removing stripe noise from satellite images as described in any of the above embodiments are implemented.

[0115] The present application also provides a computer-readable storage medium, on which a program is stored, wherein the computer-readable storage medium refers to a carrier for storing data, which may include but is not limited to a floppy disk, an optical disk, a hard disk, a flash memory, a USB flash drive, and / or a memory stick, etc., and the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The working process, working details, and technical effects of the computer-readable storage medium provided in this embodiment can be found in the above embodiment of a method for removing strip noise from satellite images, which will not be repeated here.

[0116] The application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the method for removing stripe noise from satellite images as described in any of the above embodiments.

[0117] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0118] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification. The above-mentioned embodiments only express several implementation methods of the present application, and their descriptions are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that for ordinary technicians in this field, without departing from the concept of the present application, several deformations and improvements can be made, which all belong to the protection scope of the present application. Therefore, the protection scope of the patent of this application shall be based on the attached claims.

Claims

1. A method for removing stripe noise from satellite images, characterized in that: include: Collect satellite images, and separate the satellite images into high-frequency images and low-frequency images using wavelet transform; Performing denoising on the high-frequency image to obtain a denoised high-frequency image; Splitting the low-frequency image to obtain a plurality of ground feature regions; Extracting strip noise from the feature area, obtaining image noise extraction results corresponding to each direction of the feature area, and obtaining a non-noise area and a noise area corresponding to the feature area; Performing fitting analysis on the non-noise area corresponding to the feature area and the panchromatic image, establishing a ratio relationship between the DN value of the non-noise area and the panchromatic image based on the fitting result, and correcting the noise area based on the ratio relationship and the panchromatic image to obtain a corrected image corresponding to the feature area; The corrected image is combined with the denoised high-frequency image to perform inverse wavelet transform to obtain a denoised satellite image.

2. The method for removing stripe noise from satellite images according to claim 1, characterized in that: The method of using wavelet transform to separate the satellite image into high-frequency image and low-frequency image comprises: The satellite image is separated into a low-frequency image and a high-frequency image using a Haar wavelet function, wherein the high-frequency image includes a high-frequency part image in the horizontal direction, a high-frequency part image in the vertical direction, and a high-frequency part image in the diagonal direction.

3. The method for removing stripe noise from satellite images according to claim 2, characterized in that: The denoising process is performed on the high-frequency image to obtain a denoised high-frequency image, including: The soft threshold method is used to select the threshold, and the high-frequency image is subjected to threshold processing to obtain the denoised high-frequency image. The calculation formula of the soft threshold is: Among them, w is the wavelet coefficient, T is the threshold, which is 0.5 times the image standard deviation, m is the coefficient mean, and n is the number of calculations.

4. The method for removing stripe noise from satellite images according to claim 1, characterized in that: The low-frequency image is split to obtain a plurality of ground feature areas, including: Based on the DN value of the satellite image, the low-frequency image is split into regions using a natural breakpoint method to obtain a plurality of land feature regions.

5. The method for removing stripe noise from satellite images according to claim 1, characterized in that: The step of extracting strip noise from the ground feature area includes: The Sobel operator is used to calculate the horizontal and vertical gradient values ​​of the object region, and the strip noise of the object region is extracted based on the gradient values.

6. The method for removing stripe noise from satellite images according to claim 5, characterized in that: Using the Sobel operator to calculate the horizontal and vertical gradient values ​​of the feature area, and extracting the strip noise of the feature area based on the gradient values, including: Calculate the gradient value in the horizontal direction of the feature area, obtain the area where the change of the gradient amplitude is greater than a preset threshold, and use the area as the approximate strip noise; The edge of the object is obtained based on the gradient value in the vertical direction, and the edge of the object in the approximate strip noise is removed to obtain the strip noise of the object area.

7. A device for removing strip noise from satellite images, characterized in that: include: A data acquisition module, used for acquiring satellite images, and separating the satellite images into high-frequency images and low-frequency images using wavelet transform; A high-frequency denoising module, used to perform denoising on the high-frequency image to obtain a denoised high-frequency image; A low-frequency denoising module is used to split the low-frequency image to obtain multiple ground feature areas; Extract the strip noise of the feature area, obtain the image noise extraction results corresponding to each direction of the feature area, and obtain the non-noise area and the noise area corresponding to the feature area; perform fitting analysis on the non-noise area corresponding to the feature area and the panchromatic image, establish a ratio relationship between the DN value of the non-noise area and the panchromatic image based on the fitting result, correct the noise area based on the ratio relationship and the panchromatic image, and obtain a corrected image corresponding to the feature area; The output module is used to perform inverse wavelet transform on the corrected image and the denoised high-frequency image to obtain a denoised satellite image.

8. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method for removing stripe noise from satellite images according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program, wherein when the program is executed by a processor, the method for removing stripe noise from satellite images according to any one of claims 1 to 6 is implemented.

10. A computer program product comprising computer instructions, characterized in that The computer instructions implement the steps of the method according to claims 1 to 6 when executed by a processor.

Images