Remote sensing image processing method and device, electronic equipment and medium

Abstract

The invention provides a remote sensing image processing method and device, electronic equipment and a medium, and relates to the technical field of image processing. The specific implementation scheme comprises the following steps: acquiring an original image, wherein the original image has a first dynamic range; based on the original image, a first image is obtained, the first image has a second dynamic range, and the second dynamic range is smaller than the first dynamic range; performing color balance processing on the first image to obtain a second image; and obtaining a target style image, and performing color migration processing on the second image based on the target style image to obtain a target image.

Description

Technical Field

[0001] The present invention relates to the field of computer vision and image processing technology, and in particular to a remote sensing image processing method, device, electronic equipment, storage medium and computer program product. Background Art

[0002] Due to differences in imaging conditions such as imaging time, lighting, shooting angle and ground object type, there are obvious differences in color and brightness between images. Therefore, it is necessary to perform color uniformity processing on remote sensing images to facilitate the subsequent production and use of remote sensing images. Summary of the invention

[0003] The present invention provides a remote sensing image processing method, device, electronic equipment, storage medium and computer program product.

[0004] According to one aspect of the present invention, a remote sensing image processing method is provided, including: acquiring an original image, the original image having a first dynamic range; based on the original image, acquiring a first image, the first image having a second dynamic range, the second dynamic range being smaller than the first dynamic range; performing color balancing processing on the first image to obtain a second image; acquiring a target style image, and performing color migration processing on the second image based on the target style image to obtain a target image.

[0005] According to one aspect of the present invention, a remote sensing image processing device is provided, including: a first acquisition module, used to acquire an original image, the original image has a first dynamic range; a second acquisition module, used to acquire a first image based on the original image, the first image has a second dynamic range, and the second dynamic range is smaller than the first dynamic range; a color balance module, used to perform color balance processing on the first image to obtain a second image; and a color migration module, used to acquire a target style image, and perform color migration processing on the second image based on the target style image to obtain a target image.

[0006] According to another aspect of the present invention, there is provided an electronic device, comprising: one or more processors; and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors execute the remote sensing image processing method as described above.

[0007] According to another aspect of the present invention, a computer-readable storage medium is provided, on which executable instructions are stored. When the instructions are executed by a processor, the processor executes the remote sensing image processing method as described above.

[0008] According to another aspect of the present invention, a computer program product is provided, comprising a computer program, which implements the remote sensing image processing method as described above when executed by a processor. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 is a flow chart of a remote sensing image processing method according to an embodiment of the present invention;

[0011] Figure 2 is a block diagram of a remote sensing image processing device according to an embodiment of the present invention;

[0012] Figure 3 is a block diagram of an electronic device suitable for implementing a remote sensing image processing method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0013] The technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention and the drawings in the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present invention.

[0014] It should be noted that the sequence numbers of the operations in the following method are only used as representations of the operations for the purpose of description, and should not be regarded as representing the execution order of the operations. Unless explicitly stated, the method does not need to be executed completely in the order shown.

[0015] In addition, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0016] In the technical solution of the present invention, the collection, storage, use, processing, transmission, provision, disclosure and application of the data involved (for example, including but not limited to user personal information) shall comply with the provisions of relevant laws and regulations and shall not violate public order and good morals.

[0017] In the technical solution of the present invention, before obtaining or collecting relevant data, the authorization or consent of the data owner is obtained.

[0018] Figure 1 is a flowchart of a remote sensing image processing method according to an embodiment of the present invention.

[0019] like Figure 1 As shown, the remote sensing image processing method 100 includes operations S110 to S140.

[0020] In operation S110 , an original image is acquired, where the original image has a first dynamic range.

[0021] In operation S120 , a first image is acquired based on the original image, and the first image has a second dynamic range that is smaller than the first dynamic range.

[0022] In operation S130, color balancing is performed on the first image to obtain a second image.

[0023] In operation S140 , a target style image is acquired, and a color migration process is performed on the second image based on the target style image to obtain a target image.

[0024] According to an embodiment of the present invention, the original image refers to a remote sensing image to be processed, having a first dynamic range. For example, the original image may be a remote sensing image with a high dynamic range (HDR), and each color channel of the original image may be, for example, 16 bits. The original image may include data of multiple bands, including, for example, but not limited to, red (Red, R) band data, green (Green, G) band data, blue (Blue, B) band data, and the like.

[0025] Since high dynamic range remote sensing images cannot be displayed on ordinary display devices, tone mapping technology is needed to compress the dynamic range of HDR remote sensing images and convert HDR remote sensing images into low dynamic range (LDR) remote sensing images.

[0026] In an embodiment of the present invention, for example, a local Laplacian filter may be used to process the original image to obtain a first image. The first image has a second dynamic range, which is smaller than the first dynamic range. Each color channel of the first image is, for example, 8 bits. By performing local Laplacian filtering on the original image, not only can the original image be mapped from the first dynamic range (high dynamic range) to the second dynamic range (low dynamic range), but also the edges and details in the original image can be better preserved during the dynamic range compression process without losing the bright and dark details.

[0027] After the original image is converted into the first image, the color of the first image is usually difficult to maintain a natural look. To address this problem, the first image can be subjected to color balancing processing to remove the color cast in the first image, so that the image achieves a color balance effect. In an embodiment of the present invention, a nonlinear color correction transformation method can be used to perform color balancing processing on the first image to obtain a second image with more realistic and natural colors. The process of performing color balancing processing on the first image will be described later.

[0028] According to an embodiment of the present invention, by performing local Laplace filtering and color balancing processing on the original image, a high dynamic range remote sensing image (i.e., the original image) can be mapped into a high-quality low dynamic range remote sensing image (i.e., the second image), thereby providing a good foundation for subsequent image processing and further facilitating improving the quality of the target image.

[0029] After the second image is acquired, a target style image may be acquired next, and the target style image is used as a template image to perform color migration processing on the second image to obtain the target image.

[0030] According to an embodiment of the present invention, the target style image refers to a remote sensing image with a target style. The target style includes, for example, the color and brightness of the target style image. The bit depth of each color channel of the target style image is consistent with the bit depth of each color channel of the second image, for example, both are 8 bits. The target style image can be selected according to actual needs.

[0031] In an embodiment of the present invention, by performing color migration processing on the second image using the target style image, the target style of the target style image can be migrated to the second image, so that the acquired target image has the image content of the second image and the target style of the target style image.

[0032] It can be understood that the statistical features of an image can describe the visual style of the image from an overall perspective. If two images have similar statistical features, then they will also be similar in visual style. Based on this, in the process of using the target style image to perform color migration processing on the second image, the statistical features of the target style image can be aligned with the statistical features of the second image to achieve consistency in the visual style of the two.

[0033] However, for the same image, since the features contained in the statistical features are extracted from the same data source, these features may be coupled. In the process of aligning the statistical features of the target style image with the statistical features of the second image, due to the coupling between the features, adjusting a certain feature will affect other features that have been adjusted, thereby affecting the effect of the final color migration processing.

[0034] In response to the above problems, an embodiment of the present invention performs feature decoupling processing on the statistical features of each channel of the target style image, so that the decoupled features contained in the obtained decoupled statistical features are independent of each other. Then, the second image is nested and standardized in the channel dimension, and the pixel values ​​of the corresponding channels of the second image after the nested and standardized processing are adjusted according to the decoupled statistical features corresponding to each channel of the target style image, so as to obtain the target image. In the above manner, the target image can be visually closer to the target style of the target style image, while avoiding the generation of unnatural artifacts, thereby improving the quality of the target image.

[0035] According to the technical solution of the present invention, by performing local Laplace filtering and color balancing processing on the original image, the obtained second image has more realistic and natural colors and appropriate texture edges, thereby ensuring the quality of the second image, and further using the target style image to perform color migration processing on the second image, so that the obtained target image has the image content of the second image and the target style of the target style image, thereby improving the quality of the target image and ensuring the image processing effect.

[0036] According to an embodiment of the present invention, performing color balancing processing on the first image to obtain the second image includes: acquiring preset color correction information, and performing color balancing processing on the first image based on the color correction information to obtain the second image.

[0037] The color correction information and the process of obtaining the color correction information are introduced below.

[0038] According to an embodiment of the present invention, color correction information is used to describe color distribution characteristics and color correction parameters corresponding to each of a plurality of reference images. The reference image is a color image with color cast. The color correction parameters corresponding to each reference image are used to perform color balance processing on the reference image, so that the processed reference image can achieve a color balance effect.

[0039] In one example, the color correction information includes, for example, first principal component analysis (PCA) eigenvectors and first color correction matrices corresponding to each of the plurality of reference images. The first PCA eigenvectors are used to characterize the color distribution of the corresponding reference images. The first color correction matrix is ​​a color correction parameter corresponding to the corresponding reference image, and is used to perform color balancing processing on the corresponding reference image to correct the color cast of the corresponding reference image so that the processed reference image achieves a color balance effect.

[0040] In an embodiment of the present invention, the first color correction matrix corresponding to each reference image may be acquired through the following process.

[0041] First, obtain n (n is an integer greater than 1) image pairs, each image pair includes a reference image and a target image corresponding to the reference image. The reference image is a color image with color cast, and the target image is a color image corresponding to the reference image with color balance. Both the reference image and the corresponding target image can be represented as a 3*N matrix, where N is the total number of pixels in each image.

[0042] Next, for each image pair, an objective function is constructed between the reference image and the target image included in the image pair, so as to determine the first color correction matrix corresponding to the reference image by solving the objective function.

[0043] In one example, the objective function can be expressed using the following formula (1).

[0044]

[0045] Among them, Φ([R, G, B] T )=[R, G, B, RG, RB, GB, R 2 , G 2 , B 2 , RGB, 1] T (2)

[0046] In formula (1) and formula (2), represents the i-th (i=1, 2, ..., n) reference image, represents the target image corresponding to the i-th image, M i represents the first color correction matrix corresponding to the i-th reference image, M i It can be expressed as a 3*11 matrix, F(M i ) represents the objective function of the first color correction matrix with respect to the i-th reference image, ||·|| F represents the Frobenius norm, Φ represents the mapping of the triplet [R, G, B] in the standard RGB color space (sRGB) to the high-dimensional space [R, G, B, RG, RB, GB, R 2 , B 2 , G 2 , RGB, 1] kernel function. R, G, B represent the R component, G component, and B component of each pixel in the input image in the sRGB color space. RG, RB, and GB represent the product of the R component and the G component, the R component and the B component, and the G component and the B component, respectively. 2 , G 2 and B 2They represent the square of the R component, the square of the G component, and the square of the B component respectively. RGB is the product of the R component, the G component, and the B component.

[0047] The first color correction matrix corresponding to the reference image in each image pair can be determined by the above formula (1) and formula (2). Based on the first color correction matrix, the color of the reference image can be mapped to the color of the corresponding target image to achieve color correction of the reference image, so that the processed reference image achieves a color balance effect.

[0048] In the embodiment of the present invention, the first PCA feature vector corresponding to each reference image may be obtained through the following process.

[0049] First, for each reference image, the reference image is converted from the RGB color space to the logarithmic chromaticity space, and a reference histogram feature corresponding to the reference image is constructed in the logarithmic chromaticity space.

[0050] In one example, the reference histogram features corresponding to each reference image may be determined by the following formulas (3) to (9).

[0051]

[0052] I u1(j) =log(I R(j) )-log(I G(j) ) (4)

[0053] I v1(j) =log(I R(j) )-log(I B(j) ) (5)

[0054] I u2(j) =-I u1(j) , I v2(j) =-I u1(j) +I v1(j) (6)

[0055] I u3(j) =-I v1(j) , I v3(j) =-I v1(j) +I u1(j) (7)

[0056]

[0057] In formula (3) to formula (9), j represents the jth (j=1, 2, ..., N) pixel in the reference image, and N represents the total number of pixels in the reference image. R(j) ,IG(j) and I B(j) Respectively represent the R component, G component and B component of the j-th pixel in the reference image in the RGB color space, I y(j) is the brightness component of the jth pixel in the reference image. ul(j) , l vl(j) is the u and v chrominance components of the j-th pixel in the reference image in the logarithmic chrominance space, I u2(j) ,I v2(j) ,I u3(j) ,I v3(j) It is the other chromaticity component representation of the j-th pixel in the reference image in the logarithmic chromaticity space, which is used to provide color information at different angles in order to more comprehensively describe the color distribution of the reference image. (u，v，C) Represents the histogram value of the reference image in the logarithmic chromaticity space. H(I) (u，v，C) The number of pixels in a specific color channel C, a specific u and v range is given, which helps to construct the color distribution characteristics of the reference image. Where C∈{1,2,3} represents each color channel in the histogram, ε represents the histogram bin width, and I uC(j) and l vC(j) is the chrominance component calculated according to formula (4) to formula (7). h(I) (u，v，C) Represents the reference histogram feature corresponding to the reference image. The reference histogram feature can represent the color distribution of the reference image as an m*m*3 tensor parameterized by u and v. Among them, m refers to the number of bins in the u and v dimensions in the logarithmic chromaticity space. Each bin represents a specific u and v range and is used to count the number of pixels falling within the range. m can represent the resolution or size of the histogram in the u and v dimensions.

[0058] Next, PCA processing is performed on the reference histogram features corresponding to each reference image to obtain a first PCA feature vector corresponding to the corresponding reference image.

[0059] In an embodiment of the present invention, PCA processing can be performed on the reference histogram features corresponding to the reference image so as to extract a more compact first PCA feature vector from the reference histogram features. Compared with the reference histogram features, the first PCA feature vector reduces the dimension of the data while retaining the ability to describe the color distribution. The first PCA feature vector can be used to quickly compare the similarity of color distribution between different images, which is conducive to the subsequent rapid and accurate color balance processing of the first image.

[0060] Through the above process, the first PCA eigenvector corresponding to each reference image can be determined. The first PCA eigenvector can represent the color distribution of the corresponding reference image in a more compact form.

[0061] The following is an introduction to the process of performing color balancing processing on the first image based on the above color correction information.

[0062] First, a second PCA eigenvector corresponding to the first image is determined, and the second PCA eigenvector is used to characterize the color distribution of the first image.

[0063] According to an embodiment of the present invention, the first image can be converted from the RGB color space to the logarithmic chromaticity space, and a histogram feature corresponding to the first image is constructed in the logarithmic chromaticity space, and then the histogram feature is subjected to PCA processing to obtain a second PCA feature vector corresponding to the first image. The detailed process of determining the second PCA feature vector is similar to the process of determining the first PCA feature vector corresponding to the reference image in the above embodiment, and will not be repeated here.

[0064] Next, a preset number of reference images having relatively high similarity between the first PCA feature vector and the second PCA feature vector are determined among the plurality of reference images to obtain a preset number of similar images.

[0065] It can be understood that the PCA eigenvector can characterize the color distribution of the corresponding image. If the two images are closer in color distribution, the similarity between the PCA eigenvectors corresponding to the two images is higher, and vice versa. In view of this, the similarity between the second PCA eigenvector and each first PCA eigenvector can be used to screen similar images that are more similar to the first image in color distribution, and further use the color correction matrix of these similar images to perform color balance processing on the first image to correct the color cast in the first image, thereby achieving the purpose of color balance.

[0066] In an embodiment of the present invention, for each first PCA eigenvector, the similarity between the first PCA eigenvector and the second PCA eigenvector can be determined by, for example, calculating the L2 distance between the first PCA eigenvector and the second PCA eigenvector. If the L2 distance between a first PCA eigenvector and the second PCA eigenvector is closer, it indicates that the similarity between the first PCA eigenvector and the second PCA eigenvector is higher, and vice versa.

[0067] In an embodiment of the present invention, based on the similarity ranking between the second PCA feature vector and each first PCA feature vector, reference images corresponding to a preset number of first PCA feature vectors ranked first in similarity can be selected as each similar image. It should be noted that the number of similar images can be set according to actual needs, and the present invention does not limit this.

[0068] In the embodiment of the present invention, by using the similarity between each first PCA eigenvector and the second PCA eigenvector to search for similar images corresponding to the first image, the efficiency of searching for similar images and the accuracy of similar images can be improved.

[0069] After obtaining a preset number of similar images, next, based on the first color correction matrix corresponding to each similar image in the preset number of similar images and the weight of the first color correction matrix, a second color correction matrix corresponding to the first image is constructed.

[0070] It is understandable that if two images are similar in color distribution, the color balance processing between them should also be similar. In view of this, a second color correction matrix corresponding to the first image can be constructed based on the first color correction matrix corresponding to each of the preset number of similar images screened out above and the weight of the first color correction matrix, and the second color correction matrix is ​​used to perform color balance processing on the first image.

[0071] According to an embodiment of the present invention, the weight of the first color correction matrix represents the contribution of the first color correction matrix corresponding to the corresponding similar image to the color balance processing process of the first image. It can be understood that if the first image is very similar to a similar image in color distribution, then the first color correction matrix corresponding to the similar image will have a greater contribution to the color balance processing process of the first image, and accordingly, the weight corresponding to the first color correction matrix will be larger. On the contrary, if the first image and the similar image have a large difference in color distribution, then the first color correction matrix corresponding to the similar image will have a smaller contribution to the color balance processing process of the first image, and accordingly, the weight corresponding to the first color correction matrix will be smaller. In other words, the weight corresponding to the corresponding first color correction matrix can be determined according to the similarity in color distribution between the similar image and the first image.

[0072] Based on the above content, it can be known that the L2 distance between the second PCA eigenvector and each first PCA eigenvector can be used to measure the similarity in color distribution between the first image and each similar image. Therefore, for the first color correction matrix corresponding to each similar image, the weight corresponding to the first color correction matrix can be determined according to the L2 distance between the second PCA eigenvector and the first PCA eigenvector corresponding to the corresponding similar image.

[0073] Assume that the preset number of similar images includes K similar images, and K is an integer greater than 1. In one example, the weight of the first color correction matrix can be determined by the following formula (10).

[0074]

[0075] In formula (10), αk represents the weight of the first color correction matrix corresponding to the kth (k=1, 2, ..., K) similar image, σ represents the radial attenuation factor, and dk represents the L2 distance between the second PCA eigenvector and the first PCA eigenvector corresponding to the kth similar image. For example, σ=0.25.

[0076] In one example, the following formula (11) may be used to determine the second color correction matrix corresponding to the first image.

[0077]

[0078] In formula (11), M represents the second color correction matrix corresponding to the first image, M k represents the first color correction matrix corresponding to the k-th similar image, α k Denotes the first color correction matrix M k The corresponding weight, K represents the number of similar images, k=1, 2, ..., K, K is an integer greater than 1.

[0079] After the second color correction matrix is ​​determined, next, color balancing processing is performed on the first image based on the second color correction matrix to obtain a second image.

[0080] According to an embodiment of the present invention, the second color correction matrix is ​​a nonlinear color correction matrix, which represents the color transformation from the first image with color deviation to the color balanced output image. The second color correction matrix can be used to perform nonlinear color correction transformation processing on the first image to remove the color deviation in the first image, thereby obtaining a second image with a color balanced effect.

[0081] Exemplarily, the second image may be determined using the following formulas (12) and (13).

[0082] I correct =MΦ(I input ) (12)

[0083] Among them, Φ([R, G, B] T )=[R, G, B, RG, RB, GB, R 2 , G 2 , B 2 , RGB, 1] T (13)

[0084] In formula (12) and formula (13), I correct represents the second image, M represents the second color correction matrix corresponding to the first image, and I inputrepresents the first image, Φ represents the mapping of the triplet [R, G, B] in the standard RGB color space (sRGB) to the high-dimensional space [R, G, B, RG, RB, GB, R 2 , B 2 ,G 2 ,RGB,1] kernel function.

[0085] In an embodiment of the present invention, by using the first color correction matrix corresponding to the similar image to perform color balancing processing on the first image, a second image with more realistic and natural colors can be obtained, which provides a good basis for further processing of subsequent images, thereby helping to improve the quality of the target image.

[0086] The following describes the process of performing color migration processing on the second image based on the target style image.

[0087] First, for any first pixel in the second image, a first neighborhood corresponding to the first pixel is determined in the second image, and a second neighborhood corresponding to the first neighborhood in the target style image is determined.

[0088] According to an embodiment of the present invention, for any first pixel in the second image, a first neighborhood corresponding to the first pixel is determined in the second image, and the first neighborhood is a neighborhood range centered on the first pixel. The size of the first neighborhood can be set according to actual needs and is not limited here. In an example, the size of the first neighborhood can be, for example, 100 pixels*100 pixels.

[0089] According to an embodiment of the present invention, after determining the first neighborhood, a second pixel corresponding to the first pixel can be determined in the target style image according to the coordinates of the first pixel, and a second neighborhood corresponding to the second pixel can be determined in the target style image, wherein the second neighborhood is a neighborhood range centered on the second pixel, and the second neighborhood has the same size as the first neighborhood. The second neighborhood can be used to perform color migration processing on the first neighborhood so that the visual styles of the first neighborhood and the second neighborhood are consistent.

[0090] According to an embodiment of the present invention, the second neighborhood and the first neighborhood are both local image areas including multiple channels. For example, the multiple channels may be three channels corresponding to the RGB color space, namely, the R channel, the G channel, and the B channel. When the second neighborhood is used to perform color migration processing on the first neighborhood, the color migration operation may be performed on the channel dimension, so that the pixel values ​​of each channel of the first neighborhood are properly processed and adjusted, so that the first neighborhood is visually closer to the style of the second neighborhood.

[0091] Next, the decoupling statistical features corresponding to each channel of the second neighborhood are determined.

[0092] Assume that both the first neighborhood and the second neighborhood include p (p is an integer greater than 1) channels, and the statistical features of the bth (b=1, 2, ..., p)th channel of each of the first neighborhood and the second neighborhood include the first feature, the second feature, the third feature and the fourth feature. Among them, the first feature is used to reflect the overall brightness of the bth channel of the neighborhood. The second feature is used to reflect the overall contrast of the bth channel of the neighborhood. The third feature is used to describe the asymmetry of the pixel value distribution of the bth channel of the neighborhood, which is used to reflect the overall skewness of the bth channel of the neighborhood. The fourth feature is used to describe the peak degree of the pixel value distribution of the bth channel of the neighborhood, which is used to reflect the overall kurtosis of the bth channel of the neighborhood.

[0093] In one example, the first feature can be calculated based on the average value of the pixel values ​​of all pixels in the neighborhood range in the bth channel, that is, the first feature is a first-order edge moment calculated based on the pixel values ​​of all pixels in the neighborhood range in the bth channel. The second feature can be calculated based on the average value of the squares of the pixel values ​​of all pixels in the neighborhood range in the bth channel, that is, the second feature is a second-order edge moment calculated based on the pixel values ​​of all pixels in the neighborhood range in the bth channel. The third feature can be calculated based on the average value of the cubes of the pixel values ​​of all pixels in the neighborhood range in the bth channel, that is, the third feature is a third-order edge moment calculated based on the pixel values ​​of all pixels in the neighborhood range in the bth channel. The fourth feature can be calculated based on the average value of the fourth power of the pixel values ​​of all pixels in the neighborhood range in the bth channel, that is, the fourth feature is a fourth-order edge moment calculated based on the pixel values ​​of all pixels in the neighborhood range in the bth channel.

[0094] In an embodiment of the present invention, feature decoupling processing can be performed on the statistical features of each channel of the second neighborhood based on the nested normalization method, so as to obtain the decoupled statistical features corresponding to each channel of the second neighborhood. The decoupled statistical features include a first decoupled feature, a second decoupled feature, a third decoupled feature, and a fourth decoupled feature.

[0095] The following is an introduction to the process of obtaining the decoupled statistical features corresponding to each channel of the second neighborhood.

[0096] First, for each channel of the second neighborhood, a first decoupling feature corresponding to the channel is determined according to the original pixel values ​​of all pixels within the second neighborhood in the channel.

[0097] In an embodiment of the present invention, the first feature corresponding to the bth channel of the second neighborhood can be obtained by calculating the average of the original pixel values ​​of all pixels in the second neighborhood in the bth channel, and the first decoupled feature corresponding to the bth channel can be obtained based on the first feature corresponding to the bth channel. Here, the original pixel value refers to the pixel value of the pixel in the bth channel of the second neighborhood before the feature decoupling process.

[0098] Exemplarily, the following formula (14) may be used to determine the first decoupling feature corresponding to each channel of the second neighborhood.

[0099]

[0100] In formula (14), b represents the bth channel of the second neighborhood, Xb represents the matrix composed of the original pixel values ​​of all pixels in the second neighborhood in the bth channel, μ1(X b ) represents the first feature corresponding to the bth channel of the second neighborhood, which is obtained by calculating the average of the original pixel values ​​of all pixels in the bth channel within the second neighborhood. represents the first decoupled feature corresponding to the b-th channel of the second neighborhood, where b=1, 2, ..., p. In the embodiment of the present invention, μ1(X) represents calculating the average value of all elements in X, where X is a matrix containing n_x elements.

[0101] Next, for each channel of the second neighborhood, based on the first constraint condition, a first normalization process is performed on the original pixel values ​​of all pixels within the second neighborhood in the channel to obtain a first normalization result.

[0102] According to an embodiment of the present invention, the first constraint condition includes that the first feature obtained based on the first normalization result reaches a preset first reference value. The first normalization result is used to characterize the pixel values ​​of all pixels in the second neighborhood range in the bth channel after the first normalization process.

[0103] According to an embodiment of the present invention, based on the first constraint condition, performing a first standardization processing on the original pixel values ​​of all pixels in the second neighborhood in the bth channel means that during the first standardization processing, the original pixel values ​​of all pixels in the second neighborhood in the bth channel are adjusted so that the first feature obtained based on the adjusted pixel values ​​reaches a preset first reference value.

[0104] In an embodiment of the present invention, the first normalization process can be performed on the original pixel values ​​of all pixels in the second neighborhood in the bth channel using the following formula (15).

[0105]

[0106] The first constraint can be expressed using the following formula (16).

[0107]

[0108] In formulas (15) and (16), represents the first normalization result, represents the first feature obtained based on the first normalization result, and r1 represents a preset first reference value. For example, r1 may be 0.

[0109] Next, for each channel of the second neighborhood, a second decoupling feature corresponding to the channel is determined according to the first normalization result.

[0110] In an embodiment of the present invention, the second decoupling feature corresponding to each channel of the second neighborhood can be determined by the following formula (17).

[0111]

[0112] In formula (17), represents the second decoupled feature corresponding to the bth channel of the second neighborhood, represents the second feature obtained based on the first normalization result. In the embodiment of the present invention, μ2(X) represents the average value of the squares of all elements in X.

[0113] Next, for each channel of the second neighborhood, based on the second constraint condition, a second normalization process is performed on the first normalization result to obtain a second normalization result.

[0114] According to an embodiment of the present invention, the second constraint condition includes that the first feature obtained based on the second normalization result reaches a preset first reference value, and the second feature obtained based on the second normalization result reaches a preset second reference value. The second normalization result is used to characterize the pixel values ​​of all pixels in the second neighborhood range in the bth channel after the second normalization process.

[0115] According to an embodiment of the present invention, performing a second standardization processing on the first standardization result based on the second constraint condition means that during the second standardization processing, the first feature is kept unchanged as the first reference value, and the first standardization result (that is, the pixel value of all pixels in the second neighborhood range obtained after the first standardization processing in the bth channel) is adjusted so that the first feature obtained based on the adjusted pixel value reaches the above-mentioned first reference value, and the second feature reaches the preset second reference value.

[0116] In an embodiment of the present invention, the first normalization result may be subjected to a second normalization process using the following formula (18).

[0117]

[0118] In formula (18), represents the second normalization result, Represents the second feature obtained based on the first normalization result.

[0119] The second constraint can be expressed using the following formula (19).

[0120]

[0121] In formula (19), represents the first feature obtained based on the second normalization result, represents the second feature obtained based on the second normalization result, r1 represents the preset first reference value, and r2 represents the preset second reference value. For example, r1 is 0 and r2 is 1.

[0122] Next, for each channel of the second neighborhood, a third decoupling feature corresponding to the channel is determined according to the second normalization result.

[0123] In an embodiment of the present invention, the third decoupling characteristic corresponding to each channel of the second neighborhood can be determined by the following formula (20).

[0124]

[0125] In formula (20), represents the third decoupled feature corresponding to the bth channel of the second neighborhood, represents the third feature obtained based on the second normalization result. In the embodiment of the present invention, μ3(X) represents the average value of the cubes of all elements in X.

[0126] Next, for each channel of the second neighborhood, based on the third constraint condition, according to the first standardization result and the first adjustment factor, a third standardization process is performed on the second standardization result to obtain a third standardization result.

[0127] According to an embodiment of the present invention, the third constraint condition includes that the first feature obtained based on the third standardization result reaches a preset first reference value, the second feature obtained based on the third standardization result reaches a preset second reference value, and the third feature obtained based on the third standardization result reaches a preset third reference value. The third standardization result is used to characterize the pixel values ​​of all pixels in the second neighborhood range in the bth channel after the third standardization process.

[0128] According to an embodiment of the present invention, based on the third constraint condition, performing a third standardization processing on the second standardization result according to the first standardization result and the first adjustment factor means that during the third standardization processing, the first feature is kept at the first reference value and the second feature is kept at the second reference value unchanged, and the second standardization result (that is, the pixel value of all pixels in the second neighborhood range obtained by the second standardization processing in the bth channel) is adjusted according to the first standardization result and the first adjustment factor, so that the first feature obtained based on the adjusted pixel value reaches the above-mentioned first reference value, the second feature reaches the above-mentioned second reference value, and the third feature reaches the preset third reference value.

[0129] In an embodiment of the present invention, the second normalization result is subjected to third normalization processing by the following formulas (21) to (22).

[0130]

[0131] In formulas (21) to (22), represents the third normalization result, represents the first normalization result, t0 represents the first adjustment factor, represents a first adjustment parameter, which is calculated based on the first normalization result and the first adjustment factor. represents the result of the second normalization process based on the first adjustment parameter, F(t0) represents the objective function of the first adjustment factor, represents the third-order central moment calculated based on the first adjustment parameter. and The calculation method is similar to that of , and the first adjustment parameter can be regarded as "Xb" and substituted into formula (18) to calculate and obtain the corresponding result. It means to calculate the third-order central moment of X.

[0132] The third constraint can be expressed using the following formula (23).

[0133]

[0134] In formula (23), represents the first feature obtained based on the third normalization result, represents the second feature obtained based on the third normalization result, represents the third feature obtained based on the third normalization result, r1 represents the preset first reference value, r2 represents the preset second reference value, and r3 represents the preset third reference value. For example, r1 is 0, r2 is 1, and r3 is 0.

[0135] Next, for each channel of the second neighborhood, a fourth decoupling feature corresponding to the channel is determined according to the third normalization result.

[0136] In an embodiment of the present invention, the fourth decoupling characteristic corresponding to the bth channel of the second neighborhood is determined by the following formula (24).

[0137]

[0138] In formula (24), represents the fourth decoupled feature corresponding to the b-th channel of the second neighborhood, represents the fourth feature obtained based on the third normalization result, It can be understood that the second normalization process is performed based on the first adjustment parameter to obtain an adjusted second normalization result, and the fourth feature is calculated based on the adjusted second normalization result. In an embodiment of the present invention, μ4(X) represents the calculation of the fourth-order marginal moment of X, that is, the calculation of the average value of the fourth power of all elements in X.

[0139] Through the above process, the decoupled statistical features corresponding to each channel of the second neighborhood can be obtained, including the first decoupled feature, the second decoupled feature, the third decoupled feature and the fourth decoupled feature corresponding to the channel. These decoupled features are independent of each other. In the subsequent color migration process, in each channel dimension, the corresponding visual style features of the first neighborhood can be adjusted based on each decoupled feature without affecting other visual style features, thereby making the migration process more flexible and controllable, while ensuring the quality of color migration, making the first neighborhood after color migration more natural and realistic.

[0140] Next, the original pixel values ​​of each channel of the first neighborhood are respectively subjected to nested normalization processing to obtain the target normalization result corresponding to each channel of the first neighborhood.

[0141] According to an embodiment of the present invention, the process of obtaining the target normalization result is similar to the process of obtaining the third normalization result described in the above embodiment, and will not be described again here to save space.

[0142] In an embodiment of the present invention, by performing nested normalization processing on the original pixel values ​​of each channel of the first neighborhood, a target normalization result corresponding to the corresponding channel is obtained, so that in the subsequent color migration process, while keeping certain features in the first neighborhood unchanged, other features can be independently controlled and adjusted, thereby facilitating accurate color migration processing.

[0143] Next, the corresponding target normalization result is adjusted based on the decoupling statistical features corresponding to each channel of the second neighborhood, so as to align the decoupling statistical features of each channel of the first neighborhood with the decoupling statistical features of the corresponding channel of the second neighborhood.

[0144] In an embodiment of the present invention, for each channel, the target normalization result can be adjusted by the following formulas (25) and (26).

[0145]

[0146] In formulas (25) and (26), b represents the bth channel of the second neighborhood, X s Represents the matrix consisting of the original pixel values ​​of all pixels in the first neighborhood in the bth channel, represents the decoupled statistical characteristics corresponding to the b-th channel of the second neighborhood, Represents the decoupled statistical feature pair X corresponding to the b-th channel based on the second neighborhood s The results obtained after adjustment are represents the first decoupled feature corresponding to the b-th channel of the second neighborhood, represents the second decoupled feature corresponding to the bth channel of the second neighborhood, t s represents the second adjustment factor, represents the fourth decoupled feature corresponding to the b-th channel of the second neighborhood, Indicates based on The results obtained after adjusting the target standardized results, represents the third decoupled feature corresponding to the bth channel of the second neighborhood, represents the second adjustment parameter, represents the result of the second normalization process based on the second adjustment parameter, F(t s ) represents the objective function with respect to the second adjustment factor, represents the third decoupling characteristic calculated based on the second adjustment parameter. and The calculation method is similar to and The calculation method is similar, and the second adjustment parameter can be regarded as “X b ”, and substitute them into formula (18) and formula (20) to obtain the corresponding results.

[0147] Through the above operation, the color style of the second neighborhood can be transferred to the first neighborhood. Repeat the above operation until all the first pixels in the second image are processed, and the color of the second image can be transferred based on the target style image to obtain a third image with the target style.

[0148] According to an embodiment of the present invention, before the second image is subjected to color migration processing using the target style image, the second image may be subjected to quality detection to determine whether the second image contains a first region, where the first region refers to a region in the second image where data is missing. The first region may include, for example, at least one of a noise region, a null region, a thick cloud region, a thin cloud region, a fog region, a shadow region, and a blurred region. The noise region may include, for example, a Gaussian noise region, a salt and pepper noise region, a Poisson noise region, an impulse noise region, a stripe noise region, and / or a mixed noise region. The mixed noise region refers to a region in which two or more types of noise exist in the noise region.

[0149] According to an embodiment of the present invention, a method based on machine learning or a related detection algorithm may be used to perform quality detection on the second image to determine whether the second image contains the first area. Take the first area as a thin cloud area as an example. For example, a cloud detection algorithm (such as an Fmask algorithm) or a trained convolutional neural network may be used to detect the second image to determine whether the second image contains a thin cloud area.

[0150] In some embodiments, if it is determined that the second image contains the first area, after obtaining the third image, the target mask, the second image and the third image can be fused to obtain a target image with a target style, which also fuses the image information of the first area in the second image and the image information of the area corresponding to the first area in the third image.

[0151] In other embodiments, if it is determined after quality inspection on the second image that the second image does not contain the first area, that is, there is no area with missing data in the second image, then after acquiring the third image, the third image can be determined as the target image.

[0152] The following describes the process of fusing the target mask, the second image and the third image to obtain the target image.

[0153] According to an embodiment of the present invention, a target mask is generated based on a first region in a second image. The target mask is an image mask (Mask) corresponding to the second image. The target mask can be represented, for example, by a two-dimensional matrix. The value of the element corresponding to the first region in the matrix is ​​a first value, and the value of the element corresponding to the region outside the first region is a second value. The first value and the second value are different. For example, the first value can be, for example, a value of 1, and the second value can be, for example, a value of 0.

[0154] According to an embodiment of the present invention, for example, a Laplacian Pyramid Blending algorithm may be used to perform fusion processing on the target mask, the second image, and the third image to obtain a target image with a target style.

[0155] In one example, fusing the target mask, the second image, and the third image to obtain the target image includes the following process.

[0156] First, a first Laplacian pyramid is constructed based on the second image, a second Laplacian pyramid is constructed based on the third image, and a Gaussian pyramid is constructed based on the target mask.

[0157] The process of constructing the Gaussian pyramid and the Laplacian pyramid in the embodiment of the present invention is similar to the process described in the related art, and will not be described here in order to save space.

[0158] Next, the Gaussian pyramid, the first Laplacian pyramid and the second Laplacian pyramid are fused to obtain the target image.

[0159] For example, first, according to the mask image of each layer in the Gaussian pyramid, the images of the corresponding layers in the first Laplacian pyramid and the second Laplacian pyramid are fused to obtain the fused images of the corresponding layers. Through this fusion method, a fused image from the bottom layer to the top layer can be obtained. Next, starting from the fused image of the top layer, the fused image of the next layer is superimposed after upsampling layer by layer until the fused image of the bottom layer is superimposed, thereby obtaining the target image. Specifically, starting from the fused image of the top layer, after upsampling layer by layer to the same size as the fused image of the next layer, the fused image of the next layer is superimposed, and the operation of superimposing after upsampling is repeated continuously until the fused image of the bottom layer is superimposed, and the fused image obtained after the final superposition is output as the target image.

[0160] Figure 2 It is a block diagram of a remote sensing image processing device according to an embodiment of the present invention.

[0161] like Figure 2 As shown, the remote sensing image processing device 200 includes: a first acquisition module 210 , a second acquisition module 220 , a color balance module 230 and a color migration module 240 .

[0162] The first acquisition module 210 is used to acquire an original image, and the original image has a first dynamic range.

[0163] The second acquisition module 220 is used to acquire a first image based on the original image, where the first image has a second dynamic range that is smaller than the first dynamic range.

[0164] The color balancing module 230 is used to perform color balancing processing on the first image to obtain a second image.

[0165] The color migration module 240 is used to obtain a target style image, and perform color migration processing on the second image based on the target style image to obtain a target image.

[0166] It should be noted that the implementation methods, technical problems solved, functions realized, and technical effects achieved of each module in the device part embodiment are the same or similar to the implementation methods, technical problems solved, functions realized, and technical effects achieved of each corresponding step in the method part embodiment, and will not be repeated here.

[0167] Figure 3 A block diagram of an electronic device suitable for implementing a remote sensing image processing method according to an embodiment of the present invention is schematically shown.

[0168] like Figure 3 As shown, the electronic device 300 according to an embodiment of the present invention includes a processor 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage part 308 into a random access memory (RAM) 303. The processor 301 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or a related chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 301 may also include an onboard memory for caching purposes. The processor 301 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present invention.

[0169] In RAM 303, various programs and data required for the operation of electronic device 300 are stored. Processor 301, ROM 302 and RAM 303 are connected to each other through bus 304. Processor 301 performs various operations of the method flow according to the embodiment of the present invention by executing the program in ROM 302 and / or RAM 303. It should be noted that the program can also be stored in one or more memories other than ROM 302 and RAM 303. Processor 301 can also perform various operations of the method flow according to the embodiment of the present invention by executing the program stored in the one or more memories.

[0170] According to an embodiment of the present invention, the electronic device 300 may further include an input / output (I / O) interface 305, which is also connected to the bus 304. The electronic device 300 may further include one or more of the following components connected to the I / O interface 305: an input portion 306 including a keyboard, a mouse, etc.; an output portion 307 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 308 including a hard disk, etc.; and a communication portion 309 including a network interface card such as a LAN card, a modem, etc. The communication portion 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as needed. A removable medium 311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 310 as needed, so that a computer program read therefrom is installed into the storage portion 308 as needed.

[0171] The present invention also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiment; or may exist independently without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the remote sensing image processing method according to the embodiment of the present invention is implemented.

[0172] According to an embodiment of the present invention, the computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, may include but is not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, the computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, an apparatus or a device. For example, according to an embodiment of the present invention, the computer-readable storage medium may include the ROM 302 and / or RAM 303 described above and / or one or more memories other than ROM 302 and RAM 303.

[0173] The embodiment of the present invention also includes a computer program product, which includes a computer program, and the computer program contains program code for executing the method shown in the flowchart. When the computer program product is run in a computer system, the program code is used to enable the computer system to implement the remote sensing image processing method provided by the embodiment of the present invention.

[0174] The computer program executes the above functions defined in the system / device of the embodiment of the present invention when it is executed by the processor 301. According to the embodiment of the present invention, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0175] In one embodiment, the computer program may rely on tangible storage media such as optical storage devices, magnetic storage devices, etc. In another embodiment, the computer program may also be transmitted and distributed in the form of signals on a network medium, and downloaded and installed through the communication part 309, and / or installed from the removable medium 311. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0176] In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 309, and / or installed from the removable medium 311. When the computer program is executed by the processor 301, the above functions defined in the system of the embodiment of the present invention are performed. According to the embodiment of the present invention, the system, device, means, module, unit, etc. described above can be implemented by a computer program module.

[0177] According to an embodiment of the present invention, the program code for executing the computer program provided by the embodiment of the present invention can be written in any combination of one or more programming languages, specifically, these computing programs can be implemented using high-level process and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, such as Java, C++, python, "C" language or similar programming languages. The program code can be executed completely on the user computing device, partially on the user device, partially on the remote computing device, or completely on the remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (e.g., using an Internet service provider to connect through the Internet).

[0178] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0179] It will be appreciated by those skilled in the art that the features described in the various embodiments of the present invention may be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in the present invention. In particular, without departing from the spirit and teachings of the present invention, the features described in the various embodiments of the present invention may be combined and / or combined in various ways. All of these combinations and / or combinations fall within the scope of the present invention.

[0180] The embodiments of the present invention are described above. However, these embodiments are only for the purpose of illustration, and are not intended to limit the scope of the present invention. Although each embodiment is described above, it does not mean that the measures in each embodiment cannot be used in combination advantageously. Without departing from the scope of the present invention, those skilled in the art may make various substitutions and modifications, which should all fall within the scope of the present invention.

Claims

1. A remote sensing image processing method, characterized in that: include: Acquire an original image, wherein the original image has a first dynamic range; Based on the original image, a first image is acquired, wherein the first image has a second dynamic range, and the second dynamic range is smaller than the first dynamic range; Performing color balancing processing on the first image to obtain a second image; A target style image is acquired, and color migration processing is performed on the second image based on the target style image to obtain a target image.

2. The method according to claim 1, characterized in that The performing color balancing processing on the first image to obtain the second image comprises: Get preset color correction information; Performing color balancing processing on the first image based on the color correction information to obtain the second image.

3. The method according to claim 2, characterized in that The color correction information includes first principal component analysis PCA eigenvectors and first color correction matrices corresponding to each of the plurality of reference images, wherein the first PCA eigenvectors are used to characterize the color distribution of the corresponding reference images, and the first color correction matrix is ​​used to perform color balance processing on the corresponding reference images; The performing color balancing processing on the first image based on the color correction information to obtain the second image comprises: Determine a second PCA eigenvector corresponding to the first image, where the second PCA eigenvector is used to characterize a color distribution of the first image; Determine a preset number of reference images having a high similarity between the first PCA feature vector and the second PCA feature vector among the plurality of reference images, and obtain a preset number of similar images; constructing a second color correction matrix corresponding to the first image based on a first color correction matrix corresponding to each similar image in the preset number of similar images and a weight of the first color correction matrix, wherein the weight represents a contribution degree of the first color correction matrix corresponding to the corresponding similar image to a color balance processing process of the first image; Performing color balancing processing on the first image based on the second color correction matrix to obtain the second image.

4. The method according to claim 3, characterized in that The weights of the first color correction matrix are determined in the following manner: For the first color correction matrix corresponding to each of the similar images, a weight corresponding to the first color correction matrix is ​​determined according to an L2 distance between the second PCA eigenvector and the first PCA eigenvector corresponding to the similar image.

5. The method according to claim 3, characterized in that: Determining the second PCA eigenvector corresponding to the first image includes: Converting the first image from the RGB color space to a logarithmic color space, and constructing a histogram feature corresponding to the first image in the logarithmic color space; Perform PCA processing on the histogram features to obtain a second PCA feature vector corresponding to the first image.

6. The method according to any one of claims 1 to 5, characterized in that The performing color migration processing on the second image based on the target style image to obtain the target image comprises: Performing color migration processing on the second image based on the target style image to obtain a third image; In response to determining that the second image contains a first area, a target mask, the second image and the third image are fused to obtain the target image, wherein the target mask is generated based on the first area in the second image, and the first area refers to an area in the first image where data is missing.

7. The method according to claim 6, characterized in that The performing color migration processing on the second image based on the target style image to obtain the target image further comprises: In response to determining that the second image does not include the first area, the third image is determined as the target image.

8. The method according to claim 6, characterized in that The performing color migration processing on the second image based on the target style image to obtain a third image comprises: For any first pixel in the second image, determine a first neighborhood corresponding to the first pixel in the second image, and determine a second neighborhood corresponding to the first neighborhood in the target style image; Determine a decoupling statistical feature corresponding to each channel of the second neighborhood; Performing nested normalization processing on the original pixel values ​​of each channel of the first neighborhood respectively to obtain a target normalization result corresponding to each channel of the first neighborhood; Adjusting the corresponding target normalization result based on the decoupling statistical feature corresponding to each channel of the second neighborhood to align the decoupling statistical feature of each channel of the first neighborhood with the decoupling statistical feature of the corresponding channel of the second neighborhood; The above operation is repeatedly performed until all first pixels in the second image are processed and the third image is obtained.

9. The method according to claim 8, characterized in that The decoupling statistical features include a first decoupling feature, a second decoupling feature, a third decoupling feature and a fourth decoupling feature; and determining the decoupling statistical features corresponding to each channel of the second neighborhood includes: for each channel of the second neighborhood, Determine a first decoupling feature corresponding to the channel according to original pixel values ​​of all pixels within the second neighborhood in the channel; Based on a first constraint condition, performing a first normalization process on the original pixel values ​​of all pixels within the second neighborhood in the channel to obtain a first normalization result, wherein the first constraint condition includes that a first feature obtained based on the first normalization result reaches a preset first reference value, and the first normalization result is used to characterize the pixel values ​​of all pixels within the second neighborhood in the channel after the first normalization process; Determining a second decoupling feature corresponding to the channel according to the first normalization result; Based on a second constraint condition, performing a second standardization process on the first standardization result to obtain a second standardization result, wherein the second constraint condition includes that a first feature obtained based on the second standardization result reaches a preset first reference value, and a second feature obtained based on the second standardization result reaches a preset second reference value, and the second standardization result is used to characterize the pixel values ​​of all pixels in the second neighborhood range in the channel after the second standardization process; Determining a third decoupling feature corresponding to the channel according to the second normalization result; Based on a third constraint, the second standardized result is subjected to a third standardized process according to the first standardized result and the first adjustment factor to obtain a third standardized result, wherein the third constraint includes that a first feature obtained based on the third standardized result reaches a preset first reference value, a second feature obtained based on the third standardized result reaches a preset second reference value, and a third feature obtained based on the third standardized result reaches a preset third reference value, the first adjustment factor is calculated based on the first standardized result, and the third standardized result is used to characterize the pixel values ​​of all pixels in the second neighborhood in the channel after the third standardized process; A fourth decoupling feature corresponding to the channel is determined according to the third normalization result.

10. The method according to claim 9, characterized in that The adjusting the corresponding target normalization result based on the decoupling statistical feature corresponding to each channel of the second neighborhood includes: The corresponding target normalization results are adjusted by the following formula: Wherein, b represents the bth channel of the second neighborhood, b is a positive integer less than or equal to p, p is the number of channels contained in the second neighborhood, and X s represents a matrix consisting of the original pixel values ​​of all pixels in the first neighborhood in the bth channel, represents the decoupled statistical feature corresponding to the b-th channel of the second neighborhood, represents the decoupled statistical feature pair X corresponding to the bth channel based on the second neighborhood s The results obtained after adjustment are represents the first decoupling feature corresponding to the b-th channel of the second neighborhood, represents the second decoupled feature corresponding to the bth channel of the second neighborhood, t s represents the second adjustment factor, represents the fourth decoupling feature corresponding to the b-th channel of the second neighborhood, represents the third decoupling feature corresponding to the b-th channel of the second neighborhood, Indicates based on The result obtained after adjusting the target standardized result, represents the second adjustment parameter, represents a third decoupling characteristic calculated based on the second adjustment parameter, represents the result of the second normalization process based on the second adjustment parameter, F(t s ) represents the objective function regarding the second adjustment factor.

11. The method according to claim 6, characterized in that The step of fusing the target mask, the second image and the third image to obtain the target image comprises: constructing a first Laplacian pyramid based on the second image, constructing a second Laplacian pyramid based on the third image, and constructing a Gaussian pyramid based on the target mask; The Gaussian pyramid, the first Laplacian pyramid and the second Laplacian pyramid are fused to obtain the target image.

12. The method according to any one of claims 1 to 5, characterized in that The acquiring the first image based on the original image comprises: Perform local Laplace filtering on the original image to obtain the first image.

13. A remote sensing image processing device, characterized in that: include: A first acquisition module, used for acquiring an original image, wherein the original image has a first dynamic range; A second acquisition module, configured to acquire a first image based on the original image, wherein the first image has a second dynamic range, and the second dynamic range is smaller than the first dynamic range; A color balancing module, used for performing color balancing processing on the first image to obtain a second image; The color migration module is used to obtain a target style image, and perform color migration processing on the second image based on the target style image to obtain a target image.

14. An electronic device, characterized in that: include: one or more processors; a memory for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors are caused to execute the method according to any one of claims 1 to 12.

15. A computer-readable storage medium, characterized in that: Executable instructions are stored thereon, and when the instructions are executed by a processor, the processor is caused to perform the method according to any one of claims 1 to 12.

16. A computer program product, characterized in that A computer program is included which, when executed by a processor, implements the method according to any one of claims 1 to 12.

Images