Remote sensing image fusion non-reference spectral quality evaluation method and device

Abstract

This invention relates to a referenceless spectral quality assessment method and apparatus for remote sensing image fusion. The method includes: obtaining the first-digit distribution characteristics of the fused image as a whole based on hypersphere color transformation and information statistics; obtaining the first-digit distribution characteristics predicted by standard Benford's law; performing a distribution metric on the first-digit distribution characteristics of the fused image as a whole and the first-digit distribution characteristics predicted by standard Benford's law to obtain a spectral distortion index; and assessing the referenceless spectral quality of the fused image based on the spectral distortion index. The referenceless spectral quality assessment method and apparatus of this invention are characterized by accurate assessment results and an assessment process that conforms to practical applications.

Description

Technical Field

[0001] This invention relates to the field of remote sensing image fusion technology, and in particular to a referenceless spectral quality assessment method and apparatus for remote sensing image fusion. Background Technology

[0002] High-resolution Earth observation satellites in orbit can provide remote sensing images with different spatial and spectral resolutions, such as high spatial resolution panchromatic (PAN) images, low spatial resolution multispectral (MS) images, and even lower spatial resolution hyperspectral (HS) images. Image fusion is an effective way to improve the spatial resolution of optical remote sensing images. Its purpose is to use high-resolution images to improve the spatial resolution of low-resolution images, while minimizing the distortion of spectral information in the fusion result.

[0003] Due to differences in fusion methods, the spectral quality of the fusion results varies, necessitating evaluation. Spectral quality evaluation can be categorized into subjective and objective assessments. Advanced satellite platforms or ground-based equipment can use objective evaluation values ​​of the spectral quality of the fusion results as metadata for subsequent applications such as image classification, target detection, and recognition. Based on whether a reference image is used for evaluating the fused image quality, objective quality evaluation of remote sensing image fusion can be divided into reference-based quality evaluation and referenceless quality evaluation. Reference-based quality evaluation relies on overly idealistic assumptions; in practical applications, a corresponding reference image (i.e., an image with higher spatial resolution) does not exist for the fusion result, and the implementation process of reference-based quality evaluation involves many uncertainties. Therefore, referenceless quality evaluation is preferred for remote sensing image fusion.

[0004] Among existing image fusion evaluation methods, the widely used referenceless spectral quality assessment metric is the spectral distortion metric of QNR. Spectral distortion index of FQNR in, Yes Improvements It is more accurate. However, The computation involves image degradation, and the degradation filter is typically a modulation transfer filter (MTF). However, in-orbit imaging sensors often experience changes in the MTF filter parameters due to component or instrument aging, resulting in uncertainty in the degradation filter and ultimately hindering existing image fusion evaluation methods. When used as a quality evaluation indicator without a reference, it is insufficient for evaluating the accuracy of spectral quality. Summary of the Invention

[0005] Therefore, it is necessary to provide a highly accurate method and apparatus for evaluating the spectral quality of remote sensing image fusion without reference, addressing the aforementioned technical problems.

[0006] In a first aspect, the present invention provides a referenceless spectral quality assessment method for remote sensing image fusion, comprising:

[0007] The distribution characteristics of the first digit of the fused image are obtained based on hypersphere color transformation and information statistics.

[0008] Obtain the distribution characteristics of the first digit predicted by standard Benford's law;

[0009] The spectral distortion index is obtained by measuring the distribution characteristics of the first digit of the fused image as a whole and the distribution characteristics of the first digit predicted by the standard Benford law.

[0010] The no-reference spectral quality of the fused image is evaluated based on the spectral distortion index.

[0011] In one embodiment, obtaining the first-digit distribution features of the fused image as a whole based on hypersphere color transformation and information statistics includes:

[0012] The fused image is subjected to hypersphere color transformation to obtain multiple angular components on the hypersphere surface;

[0013] Gaussian normalization is performed on each angular component to obtain multiple normalized angular components equal to the number of angular components;

[0014] Feature extraction is performed on each standardized angle component based on the statistical information of each standardized angle component in the fused image;

[0015] Calculate the distribution characteristics of the first digit of the overall fused image.

[0016] In one embodiment, feature extraction of each standardized angle component based on statistical information of each standardized angle component of the fused image includes:

[0017] Extract the first non-zero digit from left to right for each point value of the standardized angle component;

[0018] Perform probability statistics on non-zero numbers and record the probability of the same non-zero number appearing.

[0019] Arrange the non-zero numbers in ascending order, and use the first digit distribution characteristics of the 9-dimensional standardized angular components to form the probability distribution of all non-zero numbers.

[0020] In one embodiment, the distribution characteristics of the first digit of the fused image as a whole are calculated as follows:

[0021] The average of the first-digit distribution characteristics of all standardized angular components is the first-digit distribution characteristics of the fused image as a whole.

[0022] In one embodiment, the spectral distortion index is defined as The distribution feature of the first digit of the fused image as a whole is v FDD The distribution characteristic of the first digit predicted by standard Benford's law is v. BF ;

[0023] The spectral distortion index is obtained by measuring the distribution characteristics of the first digit of the fused image as a whole and the distribution characteristics of the first digit predicted by standard Benford's law:

[0024]

[0025] In the formula, f(a,b) is the similarity measurement function between vectors a and b.

[0026] In one embodiment, the similarity metric function is a symmetric KL divergence, Euclidean distance, or vector angle function.

[0027] In one embodiment, the value of each point of the angle component ranges from [0, π / 2].

[0028] In one embodiment, the non-zero digits include 1, 2, 3, 4, 5, 6, 7, 8, and 9.

[0029] In one embodiment, the probability of the same non-zero digit appearing is the ratio of the number of the same non-zero digits to the total number of standardized angular component points.

[0030] Secondly, the present invention also provides a referenceless spectral quality assessment device for remote sensing image fusion, comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0031] The distribution characteristics of the first digit of the fused image are obtained based on hypersphere color transformation and information statistics.

[0032] Obtain the distribution characteristics of the first digit predicted by standard Benford's law;

[0033] The spectral distortion index is obtained by measuring the distribution characteristics of the first digit of the fused image as a whole and the distribution characteristics of the first digit predicted by the standard Benford law.

[0034] The spectral quality of the fused image is evaluated based on the spectral distortion index.

[0035] The aforementioned referenceless spectral quality assessment method and apparatus for remote sensing image fusion includes: obtaining the first-digit distribution characteristics of the entire fused image based on hypersphere color transformation and information statistics; obtaining the first-digit distribution characteristics predicted by standard Benford's law; performing a distribution metric on the first-digit distribution characteristics of the entire fused image and the first-digit distribution characteristics predicted by standard Benford's law to obtain a spectral distortion index; and assessing the spectral quality of the fused image based on the spectral distortion index. The referenceless spectral quality assessment method for remote sensing image fusion of this invention does not involve image degradation in the calculation of the spectral distortion index, eliminating the interference of degradation filters, resulting in more accurate assessment results. Furthermore, the spectral distortion index is the difference between the first-digit distribution characteristics of the entire fused image and the first-digit distribution characteristics predicted by standard Benford's law, without requiring the assumption of scale invariance, making the assessment process more consistent with practical applications. Attached Figure Description

[0036] Figure 1 This is a flowchart illustrating a referenceless spectral quality assessment method for remote sensing image fusion provided in an embodiment of the present invention.

[0037] Figure 2 This is a flowchart illustrating a method for evaluating the spectral quality of remote sensing image fusion without reference, as described in one embodiment of the present invention.

[0038] Figure 3 This is a flowchart illustrating a method for evaluating the spectral quality of remote sensing image fusion without reference, as described in one embodiment of the present invention.

[0039] Figure 4 This is a flowchart illustrating a method for evaluating the spectral quality of remote sensing image fusion without reference, as described in one embodiment of the present invention.

[0040] Figure 5 This is a schematic diagram illustrating the principle of a referenceless spectral quality assessment method for remote sensing image fusion in one embodiment of the present invention.

[0041] Figure 6 This is a fusion result of WorldView-3 satellite images, in which... Figure 6 (a) is an MS image. Figure 6 (b) is the PAN image. Figure 6 (c) is the image fused using the IHS method. Figure 6 (d) is the image fused using the PCA method. Figure 6 (e) is the image fused using the BT method. Figure 6 (f) is the image fused using the GS method. Figure 6 (g) is the image fused using the PRACS method. Figure 6 (h) is the image fused using the MTF-GLP method;

[0042] Figure 7 This is a result of QuickBird satellite image fusion, in which... Figure 7 (a) is an MS image. Figure 7 (b) is the PAN image. Figure 7 (c) is the image fused using the IHS method. Figure 7 (d) is the image fused using the PCA method. Figure 7 (e) is the image fusion using the Brover method. Figure 7 (f) is the image fused using the GS method. Figure 7 (g) is the image fused using the PRACS method. Figure 7 (h) is the image fused using the MTF-GLP method. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0044] The most widely used no-reference quality metric currently is QNR (Quality Without Reference). QNR evaluates the spectral and spatial quality of the fused image separately, using spectral distortion metrics. Spatial distortion index This indicates that, based on this, filter-based QNR (FQNR) was further developed. FQNR utilizes spectral consistency for spectral quality assessment and designs a new spectral distortion index. Furthermore, the evaluation of spectral quality by hybrid QNR (HQNR) and regression-based QNR (RQNR) continues to use the spectral distortion index of FQNR. However, the quality scale invariance assumption of FQNR has not been fully verified, and the modulation transfer filter (MTF) used in the image degradation process has uncertainties. Furthermore, although other no-reference quality metrics exist, there are few methods for independently assessing spectral quality.

[0045] This invention utilizes Benford's law to predict the first digital distribution (FDD) characteristics of the angular components in the hypersphere color gamut to evaluate spectral quality.

[0046] In one embodiment, such as Figure 1As shown, Figure 1 This is one of the flowcharts of a referenceless spectral quality assessment method for remote sensing image fusion provided in this embodiment of the invention, including the following steps:

[0047] S101. Obtain the distribution characteristics of the first digit of the fused image based on hypersphere color transformation and information statistics.

[0048] It should be noted that the hypersphere color transform can obtain a single intensity component and multiple angular components on the hypersphere surface of the fused image. The angular components represent the spectral information of the fused image.

[0049] S102. Obtain the distribution characteristics of the first digit predicted by the standard Benford Law.

[0050] Specifically, the first-digit distribution feature predicted by the standard Benford law is a 9-dimensional first-digit distribution feature, which is the first-digit distribution (FDD) feature of the angular components in the hypersphere color domain predicted by Benford law.

[0051] Benford's Law, also known as the First Law of Mathematics, is an inherent law of numerical statistics. It states that for all natural random variables, as long as the sample space is large enough, the probability that the first digit of each sample is 1 to 9 is stable within a certain range.

[0052] S103. The distribution characteristics of the first digit of the fused image as a whole are compared with the distribution characteristics of the first digit predicted by the standard Benford law to obtain the spectral distortion index.

[0053] The first digit distribution feature of a distortion-free remote sensing image conforms to the standard Benford's Law, and using Benford's Law to perform digit distribution feature analysis can overcome the limitations of the reference image. The calculation of the spectral distortion index does not require the assumption of scale invariance, breaking through the limitations of the reference image and making the assessment of the reference-free spectral quality of fused remote sensing images more realistic.

[0054] The difference between the first digit distribution feature of the fused image and the first digit distribution feature predicted by the standard Benford law can reflect the degree of spectral distortion. The distribution metric measures the difference between the two first digit distribution features.

[0055] S104. Evaluate the no-reference spectral quality of the fused image based on the spectral distortion index.

[0056] Specifically, the smaller the value of the spectral distortion index, the better the reference-free spectral quality of the fused image.

[0057] The referenceless spectral quality assessment method for remote sensing image fusion in this embodiment uses the spectral distortion index as the assessment indicator. The calculation of the spectral distortion index does not involve image degradation, thus eliminating the interference of degradation filters and making the assessment results more accurate.

[0058] In one embodiment, such as Figure 2 As shown, Figure 2 This is one of the flowcharts illustrating a referenceless spectral quality assessment method for remote sensing image fusion provided in this embodiment of the invention. This embodiment relates to an optional method for calculating the first-digit distribution characteristics of the fused image as a whole based on hypersphere color transformation and information statistics. Based on the above embodiment, S101 includes the following steps:

[0059] S201. Perform hypersphere color transformation on the fused image to obtain multiple angular components on the hypersphere surface.

[0060] Specifically, the value range of each point of the angle component is [0, π / 2].

[0061] S202. Perform Gaussian normalization on each angle component to obtain multiple normalized angle components equal to the number of angle components.

[0062] The multiple angular components on the hypersphere obtained by hypersphere color transformation have a value range of [0, π / 2] for each pixel. Gaussian normalization can normalize the pixel value range to [0, 1], so that the first digit of the pixel value can involve all non-zero digits. In addition, the FDD characteristics of the angular components of the Gaussian normalized multi-band remote sensing image without spectral distortion conform to the standard Benford law.

[0063] S203. Based on the statistical information of each standardized angle component of the fused image, feature extraction is performed on each standardized angle component.

[0064] S204. Calculate the distribution characteristics of the first digit of the overall fused image.

[0065] In one embodiment, such as Figure 3 As shown, Figure 3 This is one of the flowcharts illustrating a referenceless spectral quality assessment method for remote sensing image fusion provided in this embodiment of the invention. This embodiment relates to an optional method for feature extraction of each standardized angular component based on the statistical information of each standardized angular component in the fused image. Based on the above embodiment, S203 includes the following steps:

[0066] S301. Extract the first non-zero digit from left to right for each point value of the standardized angle component.

[0067] Non-zero digits include 1, 2, 3, 4, 5, 6, 7, 8, and 9.

[0068] S302. Perform probability statistics on non-zero numbers and record the probability of the same non-zero number appearing.

[0069] Specifically, the probability of the same non-zero digit appearing is the ratio of the number of the same non-zero digit to the total number of standardized angular component points.

[0070] S303. Arrange the non-zero numbers in ascending order and form the first digit distribution characteristics of the 9-dimensional standardized angular components of the probability of all non-zero numbers.

[0071] In one embodiment, the distribution characteristics of the first digit of the overall fused image are calculated as follows:

[0072] The average of the first-digit distribution characteristics of all standardized angular components is the first-digit distribution characteristics of the fused image as a whole.

[0073] In an optional embodiment, the spectral distortion index is defined as The distribution feature of the first digit of the fused image as a whole is v FDD The distribution characteristic of the first digit predicted by standard Benford's law is v. BF ;

[0074] The spectral distortion index is obtained by measuring the distribution characteristics of the first digit of the fused image as a whole and the distribution characteristics of the first digit predicted by standard Benford's law:

[0075]

[0076] In the formula, f(a,b) is the similarity measurement function between vectors a and b.

[0077] In this embodiment, the similarity measurement function is preferably symmetric KL divergence, Euclidean distance, or vector angle function.

[0078] In a specific embodiment, assume that the low-resolution multi-band image M of the fused image has N bands, each band having E rows × F columns of pixels, and the high-resolution remote sensing image P has S bands, each band having H rows × W columns of pixels. The fused image is denoted as... There are N bands, and each band has H rows × W columns of pixels. For example... Figure 4 and Figure 5 As shown, Figure 4 This is a flowchart illustrating the referenceless spectral quality assessment method for remote sensing image fusion in this embodiment. Figure 5 A schematic diagram illustrating the principle of a referenceless spectral quality assessment method for remote sensing image fusion. The referenceless spectral quality assessment method for remote sensing image fusion in this embodiment includes:

[0079] (1) Perform hypersphere color transformation on the fused image of the N-band. Using formula (2), the N-1 angular components on the hypersphere can be obtained, denoted as {θ1, θ2, ... θ N-1}. Angular component θ k It is a matrix with size H rows × W columns, k = 1, 2, ..., N-1.

[0080]

[0081] (2) For each angular component θ k (θ k ∈{θ1, θ2, ... θ N-1}) Perform Gaussian normalization, and denote the N-1 normalized angular components as Each standardized angular component There are H rows × W columns, and the value range of all points is normalized to [0,1].

[0082] It should be noted that the calculation process of Gaussian normalization can be found in the reference (J, Yang, et al, “No-reference hyperspectral image quality assessment via quality-sensitive features learning,” Remote Sensing, vol.9, no.4, pp.305, 2017. Section 2.1 StatisticsFeatures in Spectral Domain formulas (1-3)), and will not be repeated here.

[0083] (3) The k-th normalized angular component of the fused image Extract the first non-zero digit from left to right of each value (for example, if the value at a point is 0.3245, the extracted digit is 3; if the value at a point is 0.0047, the extracted digit is 4). Count the extracted digits to obtain the number of digits 1-9, denoted as Q. a , a = 1, 2, ..., 9. Each Given a matrix of H rows × W columns, the probability of the digits 1-9 appearing, i.e., the frequency FDD of the first digit, is:

[0084]

[0085] For the k-th standardized angle component The first digit frequency P k (a), where a = 1, 2, ..., 9, forms the first digit distribution feature of the 9-dimensional standardized angular components, i.e., the FDD feature:

[0086]

[0087] in,[·] T Represents the transpose of a vector. For N-1 normalized angular components... FDD features The FDD features of the fused image are obtained by taking the mean value of k = 1, 2, ..., N-1, denoted as v. FDD :

[0088]

[0089] (4) Define the probability of the number b (b∈{1,2,…9}) predicted by standard Benford's law as:

[0090]

[0091] The first-digit distribution feature (FDD feature) of the 9-dimensional normalized angular components of a distortion-free remote sensing image predicted by standard Benford's law can be expressed as:

[0092]

[0093] (5) The predicted distribution v of the standard Benford law BF V with fused image FDD The differences between distributions are considered as spectral distortion, resulting in a no-reference spectral distortion index.

[0094] (6) Based on the spectral distortion index To evaluate the fused image, The smaller the value, the better the spectral quality of the fused image.

[0095] In a more specific embodiment, the no-reference spectral quality assessment method for remote sensing image fusion is illustrated using the example of no-reference spectral quality assessment through the fusion of multispectral and panchromatic images. The images used are WorldView-3 satellite images from the NBU database. The multispectral image MS has eight bands (Coastal, Blue, Green, Yellow, Red, Red Edge, NIR1, and NIR2), each with 256 rows × 256 columns of pixels. The panchromatic image PAN is a single band with 1024 rows × 1024 columns of pixels. The fused image... There are a total of 8 bands, each with 1024 rows × 1024 columns of pixels. In this embodiment, the reference-free spectral quality assessment method for remote sensing image fusion is as follows:

[0096] (1) Fusion image of 8 bands Performing a hypersphere color transformation yields seven angular components {θ1, θ2, ..., θ7} on the hypersphere surface. Each angular component θ...k (θ k The matrix ∈{θ1, θ2, ..., θ7} has 1024 rows × 1024 columns, and the value of each point in the matrix is ​​in the range of [0, π / 2].

[0097] (2) For each θ k (θ k Gaussian normalization is performed on blocks ∈{θ1, θ2, ..., θ7}, with a block size of 32×32. When calculating the local mean and standard deviation, the local window size is 7×7. The normalized output consists of 7 angular components. Each standardized angular component There are 1024 rows × 1024 columns, and the value of each point is in the range of [0,1].

[0098] (3) Use formula (3) to calculate the angle components. The first digit frequency P k (a), a = 1, 2, ..., 9. The 9-dimensional FDD features of each normalized angular component are calculated using formula (4). The FDD features of the normalized angular components of the entire fused image are obtained using formula (5). FDD Used for spectral quality assessment. The 9-dimensional FDD features predicted by standard Benford's law are obtained using equations (6)-(7). BF .

[0099] (4) Obtain the reference-free spectral distortion index using symmetric KL divergence. The specific calculation formula is as follows:

[0100]

[0101] In this embodiment, two representative no-reference spectral distortion indices are selected. and To compare and verify the spectral distortion index of this example The performance of the image fusion methods compared included IHS, PCA, GS, BT, PRACS, and MTF-GLP. All methods were performed using Matlab 2021 on an Intel Core i7 11800H CPU. The image fusion results are shown below. Figure 6 As shown, Figure 6 It can be used for subjective visual evaluation. (By...) Figure 6 It can be seen that the IHS and BT fusion methods produce very obvious spectral distortion, the PCA and GS fusion methods have moderate spectral distortion, while the PRACS and MTF-GLP fusion methods best preserve spectral information.

[0102] It should be noted that in the image fusion result, the lighter the image grayscale, the more distorted the spectrum.

[0103] Table 1 shows the comparison results of different spectral distortion indices for different fusion methods. From Table 1, we can obtain... The results differ significantly from subjective visual assessments. The spectral distortion index proposed in this invention... With FQNR The rankings are the same for GS, PRACS, and MTF-GLP methods. Furthermore, for the BT and PCA fusion method, compared to... index, The results are more consistent with subjective visual analysis.

[0104] Table 1. Numerical results of spectral quality assessment of WorldView-3 satellite dataset fusion results.

[0105]

[0106] In one specific embodiment, the difference from the embodiment described above that uses WorldView-3 satellite images from the NBU database is that the images in this embodiment use QuickBird satellite images.

[0107] In this embodiment, the image fusion result is as follows: Figure 7 As shown, by Figure 7 It can be seen that, in terms of visual effect, the MTF-GLP method produces the best spectral quality in the fused image, followed by the PRACS method. The GS and PCA fusion methods produce fused images with moderate spectral fidelity. The IHS and Brovey methods produce fused images with very obvious spectral distortion throughout, exhibiting the worst spectral fidelity.

[0108] Table 2 shows the comparison results of different spectral distortion indices for different fusion methods in this embodiment. Table 2 displays the numerical results and ranking results of the spectral distortion index for evaluating the spectral distortion of fused images generated by different fusion methods in the QuickBird dataset. Analysis of Table 2 shows that... The best evaluation result was given to the PCA fusion method, while the worst evaluation result was given to the MTF-GLP fusion method, which is clearly contrary to the visual evaluation. and The metrics yielded similar ranking results. The numerical results and ranking of the fused images for spectral distortion assessment are consistent with our visual assessment.

[0109] Table 2. Numerical results of spectral quality assessment of QuickBird satellite dataset fusion results.

[0110]

[0111] according to Figure 6 Comparative analysis with Table 1 and Figure 7 The comparative analysis with Table 2 shows that the referenceless spectral quality assessment method for remote sensing image fusion of the present invention can still achieve a relatively accurate assessment effect without relying on subjective evaluation.

[0112] Based on the same inventive concept, this application also provides a referenceless spectral quality assessment device for remote sensing image fusion, used to implement the aforementioned referenceless spectral quality assessment method for remote sensing image fusion. The solution provided by this device is similar to the implementation described in the above method. Therefore, the specific limitations in the following embodiments of the referenceless spectral quality assessment device for one or more point remote sensing image fusions can be found in the limitations of the referenceless spectral quality assessment method for remote sensing image fusion described above, and will not be repeated here.

[0113] In one embodiment, a referenceless spectral quality assessment apparatus for remote sensing image fusion includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the referenceless spectral quality assessment method for remote sensing image fusion in any of the above embodiments.

[0114] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0115] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0116] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. A reference-free spectral quality assessment method for remote sensing image fusion, characterized in that, include: The distribution characteristics of the first digit of the fused image are obtained based on hypersphere color transformation and information statistics. Obtain the distribution characteristics of the first digit predicted by standard Benford's law; The spectral distortion index is obtained by measuring the distribution characteristics of the first digit of the overall fused image and the first digit distribution characteristics predicted by the standard Benford law. The spectral quality of the fused image is evaluated based on the spectral distortion index. The distribution features of the first digit of the fused image as a whole, obtained based on hypersphere color transformation and information statistics, include: The fused image is subjected to hypersphere color transformation to obtain multiple angular components on the hypersphere surface; Gaussian normalization is performed on each angular component to obtain multiple normalized angular components equal to the number of angular components; Feature extraction is performed on each standardized angle component based on the statistical information of each standardized angle component in the fused image; Calculate the distribution characteristics of the first digit of the fused image as a whole; Feature extraction for each standardized angle component is performed based on statistical information of each standardized angle component in the fused image, including: Extract the first non-zero digit from left to right for each point value of the standardized angle component; Perform probability statistics on non-zero numbers and record the probability of the same non-zero number appearing. Arrange the non-zero numbers in ascending order, and form the first digit distribution characteristics of the 9-dimensional standardized angular components based on the probability of all non-zero numbers. The spectral distortion index is defined as The distribution characteristics of the first digit of the fused image are as follows: The distribution characteristics of the first digit predicted by standard Benford's law are as follows: ; The spectral distortion index is obtained by measuring the distribution characteristics of the first digit of the fused image as a whole and the distribution characteristics of the first digit predicted by standard Benford's law: （1） In the formula, f ( a , b ) is a vector a and b A similarity measurement function between them.

2. The reference-free spectral quality assessment method for remote sensing image fusion according to claim 1, characterized in that, The distribution characteristics of the first digit of the overall fused image are calculated as follows: The average of the first-digit distribution characteristics of all standardized angular components is the first-digit distribution characteristics of the fused image as a whole.

3. The method for referenceless spectral quality assessment of remote sensing image fusion according to claim 1, characterized in that, The similarity metric function is KL divergence, Euclidean distance, or vector angle function.

4. The reference-free spectral quality assessment method for remote sensing image fusion according to claim 1, characterized in that, The value range of each point of the angle component is [0, π / 2].

5. The reference-free spectral quality assessment method for remote sensing image fusion according to claim 1, characterized in that, The non-zero digits include 1, 2, 3, 4, 5, 6, 7, 8, and 9.

6. The method for assessing the reference-free spectral quality of remote sensing image fusion according to claim 1, characterized in that, The probability of the same non-zero digit appearing is the ratio of the number of the same non-zero digit to the total number of standardized angular component points.

7. A referenceless spectral quality assessment device for remote sensing image fusion, characterized in that, The method includes a memory and a processor, the memory storing a computer program, characterized in that the processor executes the computer program to implement the steps of the method according to any one of claims 1 to 6.

Images