Remote sensing image denoising method and system

Abstract

The invention relates to the field of images, in particular to a remote sensing image denoising method and system, and the method comprises the steps: finding a similar block set corresponding to a reference block through a block matching mode, carrying out the hard threshold filtering of all pixel points at the same position of different similar blocks in the similar block set, and carrying out the denoising of a remote sensing image. Soft threshold filtering is carried out on each similar block in the similar block set after hard threshold filtering, and after the similarity sets of all reference blocks are filtered, a preliminary estimation image is obtained through aggregation; a reference block on the preliminary estimation image is obtained, a similar block set of the reference block on the preliminary estimation image and a similar block set of the reference block on the remote sensing image are found in a block matching mode, and the weight of each position is calculated according to the similar block set on the preliminary estimation image; and weighting the frequency domain of the pixel point at each position in the similar block set on the remote sensing image by using the weight, then carrying out inverse transformation, and obtaining a denoised remote sensing image according to the similar block set on the remote sensing image subjected to inverse transformation.

Description

Technical Field

[0001] The present invention relates to the field of images, and in particular to a remote sensing image denoising method and system. Background Art

[0002] Remote sensing images are widely used in agriculture, meteorology, environmental monitoring and other fields. They are image information of the earth's surface or atmospheric environment obtained by remote sensing equipment such as satellites and drones. Such images contain rich geospatial information and spectral data, and can efficiently monitor and analyze large areas. However, remote sensing images are often interfered by various factors during the acquisition process, such as sensor noise, atmospheric scattering, equipment jitter, etc., resulting in image quality degradation. These noises will destroy the detailed information of remote sensing images, affect the authenticity and accuracy of the images, and interfere with subsequent analysis and applications. Therefore, remote sensing image denoising has become an important research topic to improve image quality and ensure analysis accuracy.

[0003] Existing remote sensing image denoising methods mainly include traditional denoising algorithms based on spatial domain and frequency domain, as well as denoising methods based on deep learning that have emerged in recent years. Spatial domain methods, such as median filtering and mean filtering, are simple and easy to implement, but they are prone to losing image edge details while removing noise. Frequency domain methods such as wavelet transform can effectively retain image detail information, but the processing speed is slow and the effect on complex noise is limited. Deep learning methods rely on a large amount of high-quality annotated data and consume a lot of computing resources. Therefore, how to develop an efficient, high-precision and adaptable remote sensing image denoising method has become a key challenge facing the current technology field. Summary of the invention

[0004] In order to enhance the denoising effect of remote sensing images, the present invention provides a remote sensing image denoising method, the method comprising the following steps: Obtain a reference block on the captured remote sensing image, and find a similar block set corresponding to the reference block by block matching. Perform hard threshold filtering on all pixel points at the same position of different similar blocks in the similar block set, and perform soft threshold filtering on each similar block in the similar block set after hard threshold filtering. After filtering the similarity sets of all reference blocks, a preliminary estimated image is obtained through aggregation. A reference block on a preliminary estimated image is obtained, and a similar block set of the reference block on the preliminary estimated image and a similar block set on the remote sensing image are found by block matching. The weight of each position is calculated according to the similar block set on the preliminary estimated image, and the frequency domain of the pixel points at the position corresponding to the weight in the similar block set on the remote sensing image is weighted by using the weight and then inversely transformed. A denoised remote sensing image is obtained according to the similar block set on the inversely transformed remote sensing image.

[0005] Preferably, the step of performing soft threshold filtering on each similar block in the similar block set after hard threshold filtering is specifically as follows: Get the hard threshold used in hard threshold filtering; For the set of similar blocks after hard threshold filtering, the gradient information of each similar block is extracted; Obtain a basic soft threshold and a hard threshold, and obtain an adjusted soft threshold corresponding to each similar block according to the gradient information, the basic soft threshold and the hard threshold; Each similar block is filtered using the soft threshold.

[0006] Preferably, the adjusted soft threshold corresponding to each similar block is obtained according to the gradient information, the basic soft threshold and the hard threshold, specifically: Calculate the difference between the hard threshold and the basic soft threshold, and obtain the weight according to the gradient information of the similar blocks; The difference is weighted by using the weight, and the sum of the weighted result and the basic soft threshold is used as the adjusted soft threshold corresponding to the similar block.

[0007] Preferably, the weight of each position is calculated based on the similar block set on the preliminary estimated image, specifically: For each position on the reference block, a pixel point set consisting of pixel points of each similar block at the position in the similar block set on the preliminary estimated image is obtained, and a weight at the position is obtained according to the pixel point set at the position; Calculating the local gradient of each position in the similar block, and calculating the average value of the local gradients at the same position in all similar blocks, and obtaining the adjustment weight based on the average value of the local gradients; The product of the weight at the same position and the adjusted weight is taken as the weight of the position.

[0008] Preferably, the weighting of the frequency domain of the pixel points at the positions corresponding to the weights in the similar block set on the remote sensing image by using the weights and then performing inverse transformation is specifically as follows: Obtaining a set of pixel points formed by weighted corresponding positions in a set of similar blocks on the remote sensing image; Transform the pixel set, multiply the transformation result and the weight, and perform inverse transformation on the multiplied result.

[0009] In addition, the present invention also provides a remote sensing image denoising system, the system comprising the following modules: The preliminary estimation module is used to obtain the reference block on the captured remote sensing image, and find the similar block set corresponding to the reference block by block matching, perform hard threshold filtering on all pixel points at the same position of different similar blocks in the similar block set, and perform soft threshold filtering on each similar block in the similar block set after hard threshold filtering. After filtering the similarity sets of all reference blocks, a preliminary estimation image is obtained through aggregation; The final estimation module is used to obtain the reference block on the preliminary estimation image, and find the similar block set of the reference block on the preliminary estimation image and the similar block set on the remote sensing image by block matching, calculate the weight of each position according to the similar block set on the preliminary estimation image, use the weight to weight the frequency domain of the pixel points at the position corresponding to the weight in the similar block set on the remote sensing image, and then perform inverse transformation, and obtain the denoised remote sensing image according to the similar block set on the inverse transformed remote sensing image.

[0010] Preferably, the step of performing soft threshold filtering on each similar block in the similar block set after hard threshold filtering is specifically as follows: Get the hard threshold used in hard threshold filtering; For the set of similar blocks after hard threshold filtering, the gradient information of each similar block is extracted; Obtain a basic soft threshold and a hard threshold, and obtain an adjusted soft threshold corresponding to each similar block according to the gradient information, the basic soft threshold and the hard threshold; Each similar block is filtered using the soft threshold.

[0011] Preferably, the adjusted soft threshold corresponding to each similar block is obtained according to the gradient information, the basic soft threshold and the hard threshold, specifically: Calculate the difference between the hard threshold and the basic soft threshold, and obtain the weight according to the gradient information of the similar blocks; The difference is weighted by using the weight, and the sum of the weighted result and the basic soft threshold is used as the adjusted soft threshold corresponding to the similar block.

[0012] Preferably, the weight of each position is calculated based on the similar block set on the preliminary estimated image, specifically: For each position on the reference block, a pixel point set consisting of pixel points of each similar block at the position in the similar block set on the preliminary estimated image is obtained, and a weight at the position is obtained according to the pixel point set at the position; Calculating the local gradient of each position in the similar block, and calculating the average value of the local gradients at the same position in all similar blocks, and obtaining the adjustment weight based on the average value of the local gradients; The product of the weight at the same position and the adjusted weight is taken as the weight of the position.

[0013] Preferably, the weighting of the frequency domain of the pixel points at the positions corresponding to the weights in the similar block set on the remote sensing image by using the weights and then performing inverse transformation is specifically as follows: Obtaining a set of pixel points formed by weighted corresponding positions in a set of similar blocks on the remote sensing image; Transform the pixel set, multiply the transformation result and the weight, and perform inverse transformation on the multiplied result.

[0014] Finally, the present invention also provides a computer program, which implements the method described in the first aspect when executed by a processor.

[0015] The present invention adaptively adjusts the soft threshold by combining the hard threshold and the gradient information of each similar block, which helps to retain more information in the detail area and perform stronger denoising in the smooth area. In addition, hard threshold filtering is first performed to remove large noise, and then the filtering effect is further optimized by adaptive soft threshold filtering to ensure that image details are retained as much as possible while removing noise. In addition, the local gradient information in the similar block set is used to calculate the local gradient average at the same position of the similar block, and the denoising weight is adaptively adjusted based on the gradient information. In this way, the detail information and the noise area in the image can be effectively distinguished, and the problem of over-smoothing in the edge and texture-rich areas can be avoided. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a flow chart of Embodiment 1; Figure 2 To adjust the weight calculation process; Figure 3 The images before and after enhancement. DETAILED DESCRIPTION

[0017] The technical solution in this application will be described below in conjunction with the accompanying drawings.

[0018] The terms "component", "module", "system", etc. used in this application are used to represent computer-related entities, hardware, firmware, a combination of hardware and software, software, or software in execution. For example, a component can be, but is not limited to, a process, a processor, an object, an executable file, an execution thread, a program and / or a computer running on a processor. By way of illustration, both applications and computing devices running on a computing device can be components. One or more components can reside in a process and / or an execution thread, and a component can be located on a computer and / or distributed between two or more computers. In addition, these components can be executed from various computer-readable media having various data structures stored thereon. Components can, for example, communicate through local and / or remote processes according to signals having one or more data packets (e.g., data from two components interacting with another component between a local system, a distributed system and / or a network, such as the Internet interacting with other systems through signals).

[0019] Figure 1 A remote sensing image denoising method is shown, the method comprising the following steps: S1, obtain the reference block on the captured remote sensing image, and find the similar block set corresponding to the reference block by block matching, perform hard threshold filtering on all pixel points at the same position of different similar blocks in the similar block set, and perform soft threshold filtering on each similar block in the similar block set after hard threshold filtering. After filtering the similarity sets of all reference blocks, a preliminary estimated image is obtained through aggregation; When denoising the remote sensing image, a reference block on the remote sensing image is obtained. In one embodiment, the smoke image is divided into a plurality of image blocks, each of which is used as a reference block. In another embodiment, a block size is set, and a moving step is set, and each time the block is moved once, a reference block is obtained.

[0020] After obtaining the reference block, a similar block set corresponding to the reference block is found by block matching. Each reference block has a similar block set. In a preferred embodiment, the similar block set is a plurality of pixel blocks with the greatest similarity to the reference block. Preferably, the similar block set includes the reference block, that is, there is a reference block in the similar block set, and there are a plurality of pixel blocks with the greatest similarity to the reference block.

[0021] For each similar block set, the similar blocks have the same size. For example, the size of all blocks in the similar block set is 7×7, and each similar block has 49 pixels. In the similar block set, there will be multiple pixels at the same position. For example, if the similar block set has 5 similar blocks, then there will be 5 similar blocks at position (2,3), which are the pixel values ​​of position (2,3) of each similar block in the similar block set. Hard threshold filtering is performed on all pixels at the same position of different similar blocks. Specifically, discrete cosine transform is used to convert them to the frequency domain, and then hard threshold transform is used to remove frequency components less than the hard threshold, and then converted to the time domain through quasi-discrete cosine transform. The above process is to filter the same position of different similar blocks, but not for each similar block, and hard threshold filtering is very effective in removing high-frequency noise, but it is easy to cause loss of image details. Different similar blocks have certain differences. In an optional embodiment of the present invention, soft threshold filtering is further performed on each similar block. Specifically, the hard threshold used in hard threshold filtering is obtained; For the set of similar blocks after hard threshold filtering, the gradient information of each similar block is extracted; Obtain a basic soft threshold and a hard threshold, and obtain an adjusted soft threshold corresponding to each similar block according to the gradient information, the basic soft threshold and the hard threshold; Each similar block is filtered using the soft threshold.

[0022] If the gradient is small, the similar blocks do not change much, and the noise accounts for a large proportion. In order to remove the noise, a larger threshold is used to enhance the denoising effect, and vice versa. In an optional embodiment, the adjusted soft threshold corresponding to each similar block is obtained according to the gradient information, the basic soft threshold and the hard threshold, specifically: Calculate the difference between the hard threshold and the basic soft threshold, and obtain the weight according to the gradient information of the similar blocks; The difference is weighted by using the weight, and the sum of the weighted result and the basic soft threshold is used as the adjusted soft threshold corresponding to the similar block.

[0023] Assume that the hard threshold is T hard , the gradient of a similar block is g, and the basic soft threshold is T soft , according to exp(-βg) to get the weight w, and according to T soft +(T hard -T soft )w obtains the adjusted soft threshold corresponding to the similar block, where β is the adjustment factor.

[0024] After all reference blocks have completed the above operations, a clustering operation will be performed to generate a preliminary estimated image. Clustering is the weighted summation or weighted average of all estimated values ​​of the same image block. In S1, each reference block corresponds to a similar block set. For each set of pixels corresponding to a fixed position in the similar block set, hard threshold filtering will be performed. For example, the similar block set of a reference block A is A1. The size of the reference block A is the same as the size of the similar blocks in the similar block set. Hard threshold filtering is used for each position. Assuming that the similar block set A1 has 6 similar blocks, the pixel set at position (1,1) is {a11, b11, c11, d11, e11}, then hard threshold filtering is performed on this pixel; similarly, hard threshold filtering is performed on the pixel set at position (1,2). After all positions of the similar blocks are hard threshold filtered, soft threshold filtering is performed on each similar block.

[0025] S2, obtaining a reference block on the preliminary estimated image, and finding a similar block set of the reference block on the preliminary estimated image and a similar block set on the remote sensing image by block matching, calculating the weight of each position according to the similar block set on the preliminary estimated image, using the weight to weight the frequency domain of the pixel points at the position corresponding to the weight in the similar block set on the remote sensing image, and then performing an inverse transform, and obtaining a denoised remote sensing image according to the similar block set on the inversely transformed remote sensing image.

[0026] After S1, a preliminary estimated image is obtained. For the reference block on the preliminary estimated image, a similar block set of the reference block on the preliminary estimated image and a similar block set on the remote sensing image are found. For example, if a reference block on the preliminary estimated image is A, a similar block set similar to the reference block A is found on the preliminary estimated image, and a similar block set similar to the reference block A is found on the remote sensing image. In one embodiment, the similar block set of the reference block on the preliminary estimated image and the similar block set on the remote sensing image are found by block matching, specifically: multiple similar blocks that are most similar to the reference block are found on the preliminary estimated image, and the multiple similar blocks found are placed in the similar block set of the reference block on the preliminary estimated image; multiple similar blocks that are most similar to the reference block are found on the remote sensing image, and the multiple similar blocks found are placed in the similar block set of the reference block on the remote sensing image. In a more detailed embodiment, the size of the similar block set on the preliminary estimated image and the size of the similar block set on the remote sensing image, that is, the number of elements, are the same. In another embodiment, the position of the reference block on the preliminary estimation image is obtained, and the reference block on the remote sensing image is determined according to the position, and a similar block set of the reference block on the remote sensing image is found by block matching.

[0027] In a preferred embodiment, the weight of each position is calculated based on the similar block set on the preliminary estimated image, specifically: For each position on the reference block, a pixel point set consisting of pixel points of each similar block at the position in the similar block set on the preliminary estimated image is obtained, and a weight at the position is obtained according to the pixel point set at the position; Calculating the local gradient of each position in the similar block, and calculating the average value of the local gradients at the same position in all similar blocks, and obtaining the adjustment weight based on the average value of the local gradients; The product of the weight at the same position and the adjusted weight is taken as the weight of the position.

[0028] In the similar block set, each similar block has the same size. For each position on the reference block, there is a corresponding position in the similar blocks of the set. For a similar block set, each similar block has a pixel at the same position, thus forming a pixel point set at this position. For example, if the position on the reference block is (2,5), then in the similar block set corresponding to the preliminary estimated image, each similar block position (2,5) has a pixel value, and thus the pixel point set corresponding to the position (2,5) is obtained. After obtaining the pixel point set at each position, the weight is obtained by the weight calculation method of the L2 norm and / or Wiener filter. In this way, all similar blocks in the similar block set have the same weight at the same position. For example, the weight of the position (2,5) in similar block 1 is the same as the weight of the position (2,5) in similar block 2.

[0029] The above process is to calculate the weight for the same position in the similar block set, but different similar blocks have different features, further calculate the local gradient of each position, calculate the average value of the local gradient of the same position in all similar blocks, and obtain the adjustment weight based on the average value of the local gradient. Specifically, for a position of a similar block in the similar block set, calculate the gradient of this position, for the same position of all similar blocks in the similar block set, calculate the average value of the gradient, and obtain the adjustment weight of this position according to the average value. In a preferred embodiment, exp(-grad) is used to calculate the adjustment weight, where grad is the average value of the gradient.

[0030] For ease of understanding, an example is given below. For the reference block at position (2,5), each similar block in the similar block set corresponding to the preliminary estimated image of the reference block has position (2,5). The pixel values ​​of all similar blocks at position (2,5) become a set of pixel points. The weight at position (2,5) can be calculated using methods such as Wiener filtering or L2 norm. Similarly, the gradients of all similar blocks at position (2,5) are calculated. If there are 5 similar blocks in the similar block set, 5 gradients will be calculated. The average of these five gradients is the average of the local gradients of the same position (2,5) in all similar blocks. The adjusted weight is then calculated based on the average. The final weight at position (2,5) is calculated using the adjusted weight and the above weight, as shown in Figure 2 shown.

[0031] After obtaining the weights, obtain the pixel point set of all similar blocks at the same position, such as (2,5), in the similar block set corresponding to the remote sensing image, and obtain the weight calculated at the same position. Perform discrete cosine transformation on the pixel point set, multiply the transformation result with the weight at this position, and perform inverse transformation on the multiplied result. Figure 3 shown.

[0032] In addition, the present invention also provides a remote sensing image denoising system, the system comprising the following modules: The preliminary estimation module is used to obtain the reference block on the captured remote sensing image, and find the similar block set corresponding to the reference block by block matching, perform hard threshold filtering on all pixel points at the same position of different similar blocks in the similar block set, and perform soft threshold filtering on each similar block in the similar block set after hard threshold filtering. After filtering the similarity sets of all reference blocks, a preliminary estimation image is obtained through aggregation; The final estimation module is used to obtain the reference block on the preliminary estimation image, and find the similar block set of the reference block on the preliminary estimation image and the similar block set on the remote sensing image by block matching, calculate the weight of each position according to the similar block set on the preliminary estimation image, use the weight to weight the frequency domain of the pixel points at the position corresponding to the weight in the similar block set on the remote sensing image, and then perform inverse transformation, and obtain the denoised remote sensing image according to the similar block set on the inverse transformed remote sensing image.

[0033] Preferably, the step of performing soft threshold filtering on each similar block in the similar block set after hard threshold filtering is specifically as follows: Get the hard threshold used in hard threshold filtering; For the set of similar blocks after hard threshold filtering, the gradient information of each similar block is extracted; Obtain a basic soft threshold and a hard threshold, and obtain an adjusted soft threshold corresponding to each similar block according to the gradient information, the basic soft threshold and the hard threshold; Each similar block is filtered using the soft threshold.

[0034] Preferably, the adjusted soft threshold corresponding to each similar block is obtained according to the gradient information, the basic soft threshold and the hard threshold, specifically: Calculate the difference between the hard threshold and the basic soft threshold, and obtain the weight according to the gradient information of the similar blocks; The difference is weighted by using the weight, and the sum of the weighted result and the basic soft threshold is used as the adjusted soft threshold corresponding to the similar block.

[0035] Preferably, the weight of each position is calculated based on the similar block set on the preliminary estimated image, specifically: For each position on the reference block, a pixel point set consisting of pixel points of each similar block at the position in the similar block set on the preliminary estimated image is obtained, and a weight at the position is obtained according to the pixel point set at the position; Calculating the local gradient of each position in the similar block, and calculating the average value of the local gradients at the same position in all similar blocks, and obtaining the adjustment weight based on the average value of the local gradients; The product of the weight at the same position and the adjusted weight is taken as the weight of the position.

[0036] Preferably, the weighting of the frequency domain of the pixel points at the positions corresponding to the weights in the similar block set on the remote sensing image by using the weights and then performing inverse transformation is specifically as follows: Obtaining a set of pixel points formed by weighted corresponding positions in a set of similar blocks on the remote sensing image; Transform the pixel set, multiply the transformation result and the weight, and perform inverse transformation on the multiplied result.

[0037] It will be appreciated by those skilled in the art that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or in combination with computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application. It will be clearly understood by those skilled in the art that for the convenience and simplicity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. In several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic, for example, the division of the units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0038] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard drives, ROM, RAM, magnetic disks, or optical disks.

Claims

1. A remote sensing image denoising method, characterized in that: The method comprises the following steps: Obtain a reference block on the captured remote sensing image, and find a similar block set corresponding to the reference block by block matching. Perform hard threshold filtering on all pixel points at the same position of different similar blocks in the similar block set, and perform soft threshold filtering on each similar block in the similar block set after hard threshold filtering. After filtering the similarity sets of all reference blocks, a preliminary estimated image is obtained through aggregation. A reference block on a preliminary estimated image is obtained, and a similar block set of the reference block on the preliminary estimated image and a similar block set on the remote sensing image are found by block matching. The weight of each position is calculated according to the similar block set on the preliminary estimated image, and the frequency domain of the pixel points at the position corresponding to the weight in the similar block set on the remote sensing image is weighted by using the weight and then inversely transformed. A denoised remote sensing image is obtained according to the similar block set on the inversely transformed remote sensing image.

2. The method according to claim 1, characterized in that The soft threshold filtering is performed on each similar block in the similar block set after the hard threshold filtering, specifically: Get the hard threshold used in hard threshold filtering; For the set of similar blocks after hard threshold filtering, the gradient information of each similar block is extracted; Obtain a basic soft threshold and a hard threshold, and obtain an adjusted soft threshold corresponding to each similar block according to the gradient information, the basic soft threshold and the hard threshold; Each similar block is filtered using the soft threshold.

3. The method according to claim 2, characterized in that The adjusted soft threshold corresponding to each similar block is obtained according to the gradient information, the basic soft threshold and the hard threshold, specifically: Calculate the difference between the hard threshold and the basic soft threshold, and obtain the weight according to the gradient information of the similar blocks; The difference is weighted by using the weight, and the sum of the weighted result and the basic soft threshold is used as the adjusted soft threshold corresponding to the similar block.

4. The method according to claim 1, characterized in that The weight of each position is calculated based on the similar block set on the preliminary estimated image, specifically: For each position on the reference block, a pixel point set consisting of pixel points of each similar block at the position in the similar block set on the preliminary estimated image is obtained, and a weight at the position is obtained according to the pixel point set at the position; Calculating the local gradient of each position in the similar block, and calculating the average value of the local gradients at the same position in all similar blocks, and obtaining the adjustment weight based on the average value of the local gradients; The product of the weight at the same position and the adjusted weight is taken as the weight of the position.

5. The method according to claim 1, characterized in that The weighting of the frequency domain of the pixel points at the positions corresponding to the weights in the similar block set on the remote sensing image by using the weights and then performing an inverse transformation is specifically as follows: Obtaining a set of pixel points formed by weighted corresponding positions in a set of similar blocks on the remote sensing image; Transform the pixel set, multiply the transformation result and the weight, and perform inverse transformation on the multiplied result.

6. A remote sensing image denoising system, characterized in that: The system includes the following modules: The preliminary estimation module is used to obtain the reference block on the captured remote sensing image, and find the similar block set corresponding to the reference block by block matching, perform hard threshold filtering on all pixel points at the same position of different similar blocks in the similar block set, and perform soft threshold filtering on each similar block in the similar block set after hard threshold filtering. After filtering the similarity sets of all reference blocks, a preliminary estimation image is obtained through aggregation; The final estimation module is used to obtain the reference block on the preliminary estimation image, and find the similar block set of the reference block on the preliminary estimation image and the similar block set on the remote sensing image by block matching, calculate the weight of each position according to the similar block set on the preliminary estimation image, use the weight to weight the frequency domain of the pixel points at the position corresponding to the weight in the similar block set on the remote sensing image, and then perform inverse transformation, and obtain the denoised remote sensing image according to the similar block set on the inverse transformed remote sensing image.

7. The system according to claim 6, characterized in that The soft threshold filtering is performed on each similar block in the similar block set after the hard threshold filtering, specifically: Get the hard threshold used in hard threshold filtering; For the set of similar blocks after hard threshold filtering, the gradient information of each similar block is extracted; Obtain a basic soft threshold and a hard threshold, and obtain an adjusted soft threshold corresponding to each similar block according to the gradient information, the basic soft threshold and the hard threshold; Each similar block is filtered using the soft threshold.

8. The system according to claim 7, characterized in that The adjusted soft threshold corresponding to each similar block is obtained according to the gradient information, the basic soft threshold and the hard threshold, specifically: Calculate the difference between the hard threshold and the basic soft threshold, and obtain the weight according to the gradient information of the similar blocks; The difference is weighted by using the weight, and the sum of the weighted result and the basic soft threshold is used as the adjusted soft threshold corresponding to the similar block.

9. The system according to claim 6, characterized in that The weight of each position is calculated based on the similar block set on the preliminary estimated image, specifically: For each position on the reference block, a pixel point set consisting of pixel points of each similar block at the position in the similar block set on the preliminary estimated image is obtained, and a weight at the position is obtained according to the pixel point set at the position; Calculating the local gradient of each position in the similar block, and calculating the average value of the local gradients at the same position in all similar blocks, and obtaining the adjustment weight based on the average value of the local gradients; The weight of the same position is the product of the weight and the adjusted weight; The weight of each position on each similar block is obtained based on the weight of each similar block and the weight of each position on the reference block.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the computer program implements the method according to any one of claims 1 to 5.

Images