Full-band image fusion method based on improved IHS transformation

Abstract

The invention discloses a full-band image fusion method based on improved IHS transformation, and the method comprises the steps: inputting a to-be-fused single-band image A and a to-be-fused multi-band image B, and determining an execution path according to the image types of the to-be-fused single-band image A and the to-be-fused multi-band image B, the execution path comprises a first path and a second path; processing the to-be-fused single-band image A and the to-be-fused multi-band image B according to the execution path to obtain a first multi-band image and a first single-band image; and taking the first single-band image as a guide image, taking the first multi-band image as an input image, performing guide filtering on the first multi-band image band by band, and processing according to a preset fusion strategy to obtain a fused image. According to the image fusion method provided by the invention, the fusion effect is ensured, and meanwhile, more information in the aspect of the wave band number can be reserved.

Description

Technical Field

[0001] The present invention relates to the technical field of image fusion, and in particular to a full-band image fusion method based on improved IHS transformation. Background Art

[0002] Image fusion refers to the process of extracting the useful information from each channel to the maximum extent from the image data collected by multiple source channels about the same target through image processing and computer technology, and finally synthesizing it into a high-quality image to improve the accuracy, reliability and utilization rate of the interpreted image, so as to form a clear, complete and accurate description of the target. At present, there are three main methods for pixel-level remote sensing fusion technology: component replacement fusion method, multi-resolution analysis fusion method and component replacement and multi-resolution analysis combined method. The component replacement fusion method changes the multi-band image data to a new expression space through an algorithm, and uses a single band to replace the intensity component of the multi-band data to achieve fusion; the multi-resolution analysis fusion method decomposes the single-band and multi-band images at multiple scales, and reconstructs the fused image after fusing the high-frequency and low-frequency information at different resolution levels according to certain rules; the fusion method combining component replacement and multi-resolution analysis is to first transform the expression space of the multi-band, decompose the intensity component of the multi-band image and the single-band image, and then use the fused intensity component to replace the intensity component of the multi-band, and finally reconstruct the processed image to obtain a fused image.

[0003] The component replacement-based fusion method has the following disadvantages: because the component replacement-based fusion method directly replaces the intensity component, it may cause image distortion, especially in some color transition areas or areas with complex terrain types; at the same time, it is sensitive to noise, and the noise may be amplified in image fusion, resulting in unclear images.

[0004] The fusion method based on multi-resolution analysis has the following disadvantages: when fusing at different resolution levels, the high-frequency information and the low-frequency information may be improperly processed in the process of combining due to the influence of the fusion rule design, which may introduce artifacts into the image; at the same time, although multi-resolution analysis cannot completely and accurately maintain the original spectral characteristics of the multispectral image, it may cause distortion of the fused image.

[0005] The method based on the combination of component replacement and multi-resolution analysis has the following disadvantages: the computational complexity of this fusion method is high. If the method is not selected properly, there may be spectral distortion problems. In the process of component replacement and fusion of information with different resolutions, the inaccurate combination of single-band image and multi-band image information will cause the distortion or loss of some spectral information. Currently, the fusion method based on IHS transformation and wavelet decomposition is commonly used in this method. However, at present, image fusion containing IHS space color transformation is only applicable to the fusion of three-band images, and the extra bands will be discarded. For projects such as image classification that require rich spectral information, the lost extra bands may lead to a decrease in classification accuracy. When processing images with more than three bands, a large amount of band information will be directly omitted, which will greatly limit the application of IHS and its related methods in actual projects. Summary of the invention

[0006] In order to solve the technical problems existing in the background technology, the present invention proposes a full-band image fusion method based on improved IHS transformation.

[0007] The present invention proposes a full-band image fusion method based on improved IHS transformation, which comprises the following steps:

[0008] S1, inputting a single-band image A and a multi-band image B to be fused, and determining an execution path according to the image types of the single-band image A and the multi-band image B to be fused, the image types including remote sensing images and non-remote sensing images, and the execution paths including a first path and a second path;

[0009] S2. Processing the single-band image A and the multi-band image B to be fused according to the execution path to obtain a first multi-band image and a first single-band image;

[0010] S3, using the first single-band image as a guide image and the first multi-band image as an input image, performing guided filtering on the first multi-band image band by band, and then processing the image according to a preset fusion strategy to obtain a fused image;

[0011] S4. Based on the image types of the single-band image A and the multi-band image B to be fused, the fused images are fused to obtain a final full-band fused image.

[0012] Preferably, step S1 specifically includes:

[0013] When the image types of the single-band image A and the multi-band image B to be fused are specifically remote sensing images, the first path is executed;

[0014] When the image types of the single-band image A and the multi-band image B to be fused are specifically non-remote sensing images, the second path is executed.

[0015] Preferably, the first path specifically uses the single-band image A and the multi-band image B to be fused directly as the first single-band image and the first multi-band image;

[0016] The second path specifically preprocesses the single-band image A and the multi-band image B to be fused to obtain a single-band image and a multi-band image with the same spatial size after registration, and uses the single-band image and the multi-band image with the same spatial size after registration as the first single-band image and the first multi-band image.

[0017] Preferably, the preprocessing specifically includes:

[0018] Perform atmospheric correction on the fused single-band image A and multi-band image B to obtain the corresponding corrected image IMG A and IMG B ;

[0019] For the corrected image IMG A and IMG B Perform image registration to obtain the registered single-band image IMGA and multi-band image IMGB;

[0020] Upsampling the registered multi-band image IMGB to the same spatial size as the single-band image IMGA to obtain a first single-band image;

[0021] The registered multi-band image IMGB is used as the first multi-band image.

[0022] Preferably, the preset fusion strategy specifically includes:

[0023] The first single-band image is used as a guide image and the first multi-band image is used as an input image, and guided filtering is performed band by band on the first multi-band image;

[0024] When the number of bands of the first multi-band image is greater than 3, the first three bands of the first multi-band image after guided filtering are used as three-channel images, and the remaining bands of the first multi-band image after guided filtering except the first three bands are used for cascading to obtain a final full-band fusion image;

[0025] When the number of bands of the first multi-band image is equal to 3, the first multi-band image after guided filtering is used as a three-channel image;

[0026] Performing IHS forward transformation on the three-channel image to obtain an intensity component I, a chromaticity component H, and a saturation component S, and performing histogram matching on the first single-band image and the intensity component I to obtain a second single-band image;

[0027] The intensity component I and the second single-band image are processed according to a preset strategy to obtain a fused intensity component I';

[0028] The fused intensity component I', chroma component H, and saturation component S are inversely transformed by IHS to convert them into RGB color space to obtain a fused image.

[0029] Preferably, the guided filtering formula is as follows:

[0030] q = ∑ W (i) p;

[0031] Among them, i is the guide map; p is the input map; q is the three-channel image, and W is the filter window determined according to the guide map.

[0032] Preferably, the preset strategy is specifically:

[0033] Perform multi-scale decomposition on the intensity component I and the second single-band image respectively to obtain corresponding high-frequency layers and low-frequency layers;

[0034] The high-frequency layers corresponding to the intensity component I and the second single-band image are fused according to a high-frequency fusion formula to obtain a fused high-frequency layer;

[0035] The intensity component I and the low-frequency layer corresponding to the second single-band image are fused according to the low-frequency fusion formula to obtain a fused low-frequency layer;

[0036] Perform multi-scale inverse transformation on the fused high-frequency layer and the fused low-frequency layer to obtain the fused intensity component I';

[0037] The high frequency fusion formula is specifically:

[0038] PH=max(|A h |,|B h |);

[0039] Among them, PH is the high-frequency layer after fusion; A h is the high-frequency layer corresponding to the second single-band image; B h is the high-frequency layer corresponding to the intensity component I;

[0040] The low-frequency fusion formula is specifically:

[0041] PD=λ·A d +(1-λ)·B d ;

[0042] Among them, PD is the low-frequency layer after fusion; λ is the fusion coefficient, λ = 0.5; A d is the low-frequency layer corresponding to the second single-band image; B d is the low-frequency layer corresponding to the intensity component I.

[0043] Preferably, step S4 specifically includes:

[0044] When the image types of the single-band image A and the multi-band image B to be fused are specifically non-remote sensing images, the fused image is used as the final full-band fused image;

[0045] When the image types of the single-band image A and the multi-band image B to be fused are specifically remote sensing images, full-band fusion processing is performed on the fused images to obtain a final full-band fused image.

[0046] Preferably, the full-band fusion processing of the fused image to obtain the final full-band fused image specifically includes: cascading the remaining bands of the first multi-band image after guided filtering except the first three bands with the fused image in the order of the band range of the first multi-band image to obtain the final full-band fused image.

[0047] Preferably, it also includes:

[0048] The quality of the final full-band fusion image is evaluated by using a fusion image quality evaluation index, wherein the fusion image quality evaluation index includes a spectral angle evaluation index and a peak signal-to-noise ratio evaluation index;

[0049] The spectral angle evaluation index formula is as follows:

[0050]

[0051] Among them, SAM is the spectral angular distance between the two input images; ||||2 is the 2-norm operation, T represents the transposition operation; F is the original multi-band data; F f is the fused multi-band data;

[0052] The peak signal-to-noise ratio evaluation index formula is as follows:

[0053]

[0054] Among them, PSNR is the peak signal-to-noise ratio; MAX I is the maximum pixel value of the input image; MSE is the mean square error, which reflects the difference between the two input images; z represents the zth band of the multi-band image; c is the number of bands of the multi-band image, and M represents the original single-band image.

[0055] In the present invention, a full-band image fusion method based on improved IHS transformation is proposed, a single-band image A and a multi-band image B to be fused are input, and an execution path is determined according to the image types of the single-band image A and the multi-band image B to be fused, the image types include remote sensing images and non-remote sensing images, and the execution path includes a first path and a second path; the single-band image A and the multi-band image B to be fused are processed according to the execution path to obtain a first multi-band image and a first single-band image; the first single-band image is used as a guide map and the first multi-band image is used as an input map, and the first multi-band image is guided and filtered band by band, and then processed according to a preset fusion strategy to obtain a fused image; based on the image types of the single-band image A and the multi-band image B to be fused, the fused image is fused to obtain a final full-band fused image. The problem of insufficient reconstruction of spatial detail information and spectral color information in IHS-based and related image fusion methods is fully improved. In addition, the method proposed in the present invention can retain the band range information in the original multi-band image to the greatest extent. The method proposed in the present invention provides a new development direction for the traditional IHS transformation method and has a good fusion effect on both visible light and remote sensing images. For multi-band images with more than 3 bands, more information on the number of bands can be retained while ensuring the fusion effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 A schematic diagram of the workflow structure of a full-band image fusion method based on improved IHS transformation proposed by the present invention;

[0057] Figure 2 A schematic diagram of the implementation process of a full-band image fusion method based on improved IHS transformation proposed by the present invention;

[0058] Figure 3 The schematic diagram of the structure of the guided filtering process of the full-band image fusion method based on the improved IHS transform proposed by the present invention. DETAILED DESCRIPTION

[0059] Reference Figure 1-3 The present invention proposes a full-band image fusion method based on improved IHS transformation, which includes the following steps:

[0060] S1. Input a single-band image A and a multi-band image B to be fused, and determine an execution path according to the image types of the single-band image A and the multi-band image B to be fused. The image types include remote sensing images and non-remote sensing images. The execution paths include a first path and a second path.

[0061] It should be noted that for remote sensing images, the spatial resolution of single-band images is higher than that of multi-band images.

[0062] In this embodiment, step S1 specifically includes:

[0063] When the image types of the single-band image A and the multi-band image B to be fused are specifically remote sensing images, the first path is executed; when the image types of the single-band image A and the multi-band image B to be fused are specifically non-remote sensing images, the second path is executed.

[0064] In this embodiment, the first path specifically uses the single-band image A and the multi-band image B to be fused directly as the first single-band image and the first multi-band image.

[0065] The second path is specifically to preprocess the single-band image A and the multi-band image B to be fused to obtain a single-band image and a multi-band image with the same spatial size after registration, and use the single-band image and the multi-band image with the same spatial size after registration as the first single-band image and the first multi-band image.

[0066] In this embodiment, the preprocessing specifically includes: performing an atmospheric correction operation on the single-band image A and the multi-band image B to be fused to obtain a corresponding corrected image IMG A and IMG B ; For the corrected image IMG A and IMG B Perform image registration to obtain a registered single-band image IMGA and a multi-band image IMGB; upsample the registered multi-band image IMGB to the same spatial size as the single-band image IMGA to obtain a first single-band image; and use the registered multi-band image IMGB as the first multi-band image.

[0067] In this embodiment, ENVI software is used to perform atmospheric correction on remote sensing images. The purpose of this step is to obtain the real reflectance data of the surface. Before performing atmospheric correction, the remote sensing image storage format is obtained by viewing the remote sensing image header file. If it is in BSQ format, it is necessary to convert the storage format to BIL or BIP storage method through the Convert Interleave tool before performing atmospheric correction. If it is not in BSQ format, the FLAASH atmospheric correction tool is directly used for processing. The atmospheric correction process can be expressed by the following formula:

[0068] IMG j =FLAASH(j,Lat,Lon,SenH,GE,Pixel,M,Y,D,T,Atmo,Aero),j=A,B;

[0069] Where j is the image to be corrected, IMG jIt represents the corrected image, Lat and Lon represent the latitude and longitude of the center point of the image, SenH represents the sensor height, GE represents the average elevation of the image area (obtained through the Compute BandStatistics tool in ENVI), Pixel represents the image spatial resolution, M, Y, D, T represent the image imaging month, year, day and time, Atmo and Aero represent the selected atmospheric model and aerosol model (determined from the official document through the imaging time of the specific image). The above parameters, except GE, Atmo and Aero, can be obtained from metadata.

[0070] In this embodiment, ENVI is used to register single-band and multi-band remote sensing images. The purpose of this step is to ensure the accuracy of the fusion process.

[0071] The pixel-level fusion method based on IHS transformation has strict requirements on image registration. The higher the image registration accuracy, the better the fusion image effect. Image registration is performed on the atmospherically corrected image IMG through the Image to Image tool of ENVI. The image registration process can be expressed by the following formula:

[0072] IMGA,IMGB=Registration(IMG A ,IMG B );

[0073] Among them, IMGA represents the single-band image after registration; IMGB represents the multi-band image after registration; IMG A Represents the corrected image of the single-band image A to be fused; IMG B Represents the corrected image of the multi-band image B to be fused.

[0074] In this embodiment, the registered multi-band image IMGB is upsampled to the same spatial size as the single-band image IMGA according to the following formula. The specific formula is as follows:

[0075] IMGB=resize(IMGB,ratiox,ratioy);

[0076] Among them, retiox and ratioy represent the aspect ratios of the single-band image A and the multi-band image B respectively.

[0077] S2. Processing the single-band image A and the multi-band image B to be fused according to the execution path to obtain a first multi-band image and a first single-band image.

[0078] S3, using the first single-band image as a guide image and the first multi-band image as an input image, performing guided filtering on the first multi-band image band by band, and then processing the image according to a preset fusion strategy to obtain a fused image.

[0079] In this embodiment, the preset fusion strategy specifically includes: taking the first single-band image as the guide image and the first multi-band image as the input image, and performing guided filtering on the first multi-band image band by band; when the number of bands of the first multi-band image is greater than 3, taking the first three bands of the first multi-band image after guided filtering as a three-channel image, and using the remaining bands of the first multi-band image after guided filtering except the first three bands for cascading to obtain a final full-band fused image; when the number of bands of the first multi-band image is equal to 3, taking the first multi-band image after guided filtering as a three-channel image; performing IHS forward transformation on the three-channel image to obtain an intensity component I, a chromaticity component H, and a saturation component S, and performing histogram matching on the first single-band image and the intensity component I to obtain a second single-band image; processing the intensity component I and the second single-band image according to the preset strategy to obtain a fused intensity component I'; performing IHS inverse transformation on the fused intensity component I', chromaticity component H, and saturation component S to convert to RGB color space to obtain a fused image.

[0080] In this embodiment, an IHS forward transformation is performed on the three-channel image to obtain an intensity component I, a chromaticity component H, and a saturation component S. The method of converting the RGB color space to the IHS space is as follows:

[0081]

[0082] In this embodiment, the guided filtering formula is as follows:

[0083] q = ∑ W (i) p;

[0084] Among them, i is the guide map; p is the input map; q is the three-channel image, and W is the filter window determined according to the guide map.

[0085] In this embodiment, the preset strategy is specifically as follows: performing multi-scale decomposition on the intensity component I and the second single-band image respectively to obtain corresponding high-frequency layers and low-frequency layers; fusing the high-frequency layers corresponding to the intensity component I and the second single-band image according to the high-frequency fusion formula to obtain a fused high-frequency layer; fusing the low-frequency layers corresponding to the intensity component I and the second single-band image according to the low-frequency fusion formula to obtain a fused low-frequency layer; performing multi-scale inverse transformation on the fused high-frequency layer and the fused low-frequency layer to obtain a fused intensity component I'.

[0086] In this embodiment, the high frequency fusion formula is specifically:

[0087] PH=max(|Ah |,|B h |);

[0088] Among them, PH is the high-frequency layer after fusion; A h is the high-frequency layer corresponding to the second single-band image; B h is the high-frequency layer corresponding to the intensity component I;

[0089] The low-frequency fusion formula is as follows:

[0090] PD=λ·A d +(1-λ)·B d ;

[0091] Among them, PD is the low-frequency layer after fusion; λ is the fusion coefficient, λ = 0.5; A d is the low-frequency layer corresponding to the second single-band image; B d is the low-frequency layer corresponding to the intensity component I.

[0092] In this embodiment, the fused intensity component I', chroma component H, and saturation component S are subjected to an IHS inverse transformation to convert to an RGB color space to obtain a fused image. The method of converting the IHS color space to the RGB space is as follows:

[0093]

[0094] S4. Based on the image types of the single-band image A and the multi-band image B to be fused, the fused images are fused to obtain a final full-band fused image.

[0095] In this embodiment, step S4 specifically includes:

[0096] When the image types of the single-band image A and the multi-band image B to be fused are specifically non-remote sensing images, the fused image is used as the final full-band fused image; when the image types of the single-band image A and the multi-band image B to be fused are specifically remote sensing images, the fused image is subjected to full-band fusion processing to obtain the final full-band fused image.

[0097] In this embodiment, the fused image is subjected to full-band fusion processing to obtain a final full-band fused image, specifically comprising: cascading the remaining bands of the first multi-band image after guided filtering except the first three bands with the fused image in the order of the band range of the first multi-band image to obtain the final full-band fused image.

[0098] In this embodiment, it also includes: using a fusion image quality evaluation index to evaluate the quality of the final full-band fusion image, the fusion image quality evaluation index includes a spectral angle evaluation index and a peak signal-to-noise ratio evaluation index;

[0099] The spectral angle evaluation index formula is as follows:

[0100]

[0101] Among them, SAM is the spectral angular distance between the two input images; ||||2 is the 2-norm operation, T represents the transposition operation; F is the original multi-band data; F f is the fused multi-band data;

[0102] The peak signal-to-noise ratio evaluation index formula is as follows:

[0103]

[0104] Among them, PSNR is the peak signal-to-noise ratio; MAX I is the maximum pixel value of the input image; MSE is the mean square error, which reflects the difference between the two input images; z represents the zth band of the multi-band image; c is the number of bands of the multi-band image, and M represents the original single-band image.

[0105] In the specific working process of the full-band image fusion method based on the improved IHS transform of this embodiment, first, it is determined whether the image is a remote sensing image and the image preprocessing is completed. Then, the single-band image is used as the guide image and the multi-band image is used as the input image. The multi-band image is guided and filtered band by band, and the number of bands of the multi-band image is determined. If it is greater than three bands, the first three bands of the multi-band image are decomposed by IHS and the intensity component I and the single-band image are decomposed by a multi-scale method. Then, a new intensity component is obtained according to the fusion rule, and then an inverse transformation is performed to obtain a three-band fused image. Finally, it is cascaded with the remaining bands in the channel dimension to achieve high-quality, high-efficiency, and full-band fusion of the two input images. If it is a three-band image, the IHS transform is directly performed, and the I component and the single-band image are decomposed by a multi-scale method and then fused according to the fusion rule.

[0106] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A full-band image fusion method based on improved IHS transform, characterized in that: The following steps are involved: S1, inputting a single-band image A and a multi-band image B to be fused, and determining an execution path according to the image types of the single-band image A and the multi-band image B to be fused, the image types including remote sensing images and non-remote sensing images, and the execution paths including a first path and a second path; S2. Processing the single-band image A and the multi-band image B to be fused according to the execution path to obtain a first multi-band image and a first single-band image; S3, using the first single-band image as a guide image and the first multi-band image as an input image, performing guided filtering on the first multi-band image band by band, and then processing the image according to a preset fusion strategy to obtain a fused image; S4. Based on the image types of the single-band image A and the multi-band image B to be fused, the fused image is fused to obtain a final full-band fused image.

2. The full-band image fusion method based on improved IHS transformation according to claim 1, characterized in that: Step S1 specifically includes: When the image types of the single-band image A and the multi-band image B to be fused are specifically remote sensing images, the first path is executed; When the image types of the single-band image A and the multi-band image B to be fused are specifically non-remote sensing images, the second path is executed.

3. The full-band image fusion method based on improved IHS transformation according to claim 2 is characterized in that: The first path specifically uses the single-band image A and the multi-band image B to be fused directly as the first single-band image and the first multi-band image; The second path specifically preprocesses the single-band image A and the multi-band image B to be fused to obtain a single-band image and a multi-band image with the same spatial size after registration, and uses the single-band image and the multi-band image with the same spatial size after registration as the first single-band image and the first multi-band image.

4. The full-band image fusion method based on improved IHS transformation according to claim 3 is characterized in that: The pre-processing specifically includes: Perform atmospheric correction on the fused single-band image A and multi-band image B to obtain the corresponding corrected image IMG A and IMG B ; For the corrected image IMG A and IMG B Perform image registration to obtain the registered single-band image IMGA and multi-band image IMGB; Upsampling the registered multi-band image IMGB to the same spatial size as the single-band image IMGA to obtain a first single-band image; The registered multi-band image IMGB is used as the first multi-band image.

5. The full-band image fusion method based on improved IHS transformation according to claim 1, characterized in that: The preset fusion strategy specifically includes: The first single-band image is used as a guide image and the first multi-band image is used as an input image, and guided filtering is performed band by band on the first multi-band image; When the number of bands of the first multi-band image is greater than 3, the first three bands of the first multi-band image after guided filtering are used as three-channel images, and the remaining bands of the first multi-band image after guided filtering except the first three bands are used for cascading to obtain a final full-band fusion image; When the number of bands of the first multi-band image is equal to 3, the first multi-band image after guided filtering is used as a three-channel image; Performing IHS forward transformation on the three-channel image to obtain an intensity component I, a chromaticity component H, and a saturation component S, and performing histogram matching on the first single-band image and the intensity component I to obtain a second single-band image; The intensity component I and the second single-band image are processed according to a preset strategy to obtain a fused intensity component I'; The fused intensity component I', chroma component H, and saturation component S are inversely transformed by IHS to convert them into RGB color space to obtain a fused image.

6. The full-band image fusion method based on improved IHS transformation according to claim 5, characterized in that: The guided filtering formula is as follows: q = ∑ W (i) p; Among them, i is the guide map; p is the input map; q is the three-channel image, and W is the filter window determined according to the guide map.

7. The full-band image fusion method based on improved IHS transformation according to claim 5, characterized in that: The preset strategy is specifically: Perform multi-scale decomposition on the intensity component I and the second single-band image respectively to obtain corresponding high-frequency layers and low-frequency layers; The high-frequency layers corresponding to the intensity component I and the second single-band image are fused according to a high-frequency fusion formula to obtain a fused high-frequency layer; The intensity component I and the low-frequency layer corresponding to the second single-band image are fused according to the low-frequency fusion formula to obtain a fused low-frequency layer; Perform multi-scale inverse transformation on the fused high-frequency layer and the fused low-frequency layer to obtain the fused intensity component I'; The high frequency fusion formula is specifically: PH=max(A h |,|B h ); Among them, PH is the high-frequency layer after fusion; A h is the high-frequency layer corresponding to the second single-band image; B h is the high-frequency layer corresponding to the intensity component I; The low-frequency fusion formula is specifically: PD=λ·A d +(1-λ)·B d； Among them, PD is the low-frequency layer after fusion; λ is the fusion coefficient, λ = 0.5; A d is the low-frequency layer corresponding to the second single-band image; B d is the low-frequency layer corresponding to the intensity component I.

8. The full-band image fusion method based on improved IHS transformation according to claim 5, characterized in that: Step S4 specifically includes: When the image types of the single-band image A and the multi-band image B to be fused are specifically non-remote sensing images, the fused image is used as the final full-band fused image; When the image types of the single-band image A and the multi-band image B to be fused are specifically remote sensing images, full-band fusion processing is performed on the fused images to obtain a final full-band fused image.

9. The full-band image fusion method based on improved IHS transformation according to claim 8, characterized in that: The full-band fusion processing is performed on the fused image to obtain the final full-band fused image, specifically comprising: cascading the remaining bands of the first multi-band image after the guided filtering except the first three bands with the fused image in the order of the band range of the first multi-band image to obtain the final full-band fused image.

10. The full-band image fusion method based on improved IHS transformation according to claim 1, characterized in that: Also includes: The quality of the final full-band fusion image is evaluated by using a fusion image quality evaluation index, wherein the fusion image quality evaluation index includes a spectral angle evaluation index and a peak signal-to-noise ratio evaluation index; The spectral angle evaluation index formula is as follows: Where SAM is the spectral angular distance between the two input images; || ||2 is the 2-norm operation, T represents the transposition operation; F is the original multi-band data; F f is the fused multi-band data; The peak signal-to-noise ratio evaluation index formula is as follows: Among them, PSNR is the peak signal-to-noise ratio; MAX I is the maximum pixel value of the input image; MSE is the mean square error, which reflects the difference between the two input images; z represents the zth band of the multi-band image; c is the number of bands of the multi-band image, and M represents the original single-band image.

Images