Multiband image fusion method and device based on significant features

Through the methods of significant feature extraction and image layering, the problem of low image fusion effect and efficiency in multi-band infrared detection technology is solved, and the efficient fusion of visible light, short-wave infrared and long-wave infrared images is achieved, which is suitable for the detection requirements of modern airports , improves the pilot's situational awareness ability.

CN118982470APending Publication Date: 2024-11-19BEIJING AERONAUTIC SCI & TECH RES INST OF COMAC +1
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411201476.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-29
Publication Date
2024-11-19

AI Technical Summary

Technical Problem

When multi-band infrared detection technology fuses multiple band images, it is difficult to effectively retain the characteristic information of each band, and the fusion efficiency is low. Especially when processing visible light, short-wave infrared and long-wave infrared images, the existing technology cannot meet modern airport lighting. detection requirements.

Method used

Using a multi-band image fusion method based on distinctive features, the image layering of multi-band images is performed, the significance features are extracted, the fusion coefficient matrix is ​​calculated, and the weighted fusion of the detail layer and the basic layer is carried out to obtain the final fusion result.

Benefits of technology

Improves the fusion effect and efficiency of multi-band images, and can reduce the fusion time while retaining the features of each band while achieving low delay and rich features. It is suitable for enhancing the visual equipment and improving pilots at night and low visibility scenarios. situational awareness ability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118982470A_ABST
    Figure CN118982470A_ABST
Patent Text Reader

Abstract

The embodiment of the invention discloses a multi-band image fusion method based on significant features, and the method comprises the steps: determining a basic layer image and a detail layer image of any target band image in a multi-band image group; obtaining the saliency characteristics of each target band image in the multi-band image group, and obtaining a fusion coefficient matrix corresponding to the detail layer image of each target band image by using the saliency characteristics of each target band image; fusing the detail layer images of the target wave band images by using the fusion coefficient matrix corresponding to the detail layer images of the target wave band images to obtain a detail layer fusion result; and fusing the base layer images of the target wave band images to obtain a base layer fusion result, and according to the basic layer fusion result and the detail layer fusion result, obtaining a fusion result of each target wave band image.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a multi-band image fusion method and device based on significant features. Background Art

[0002] With the continuous development of infrared detection technology, the detection performance in weak light environments has been continuously improved, and the detectable infrared light bands have been increasing, which has also promoted the application of infrared detection technology in more fields. For example, for the enhanced vision system (EVS), the next generation of EVS is moving towards multi-band infrared detection in order to provide pilots with more abundant external visual data of the aircraft. Most of the existing EVS equipment is based on a single infrared band architecture and lacks a visible light imaging band. In the process of civil aviation airport construction, traditional airport navigation lights with halogen lamps, metal halides, and bromine tungsten as light sources are gradually replaced by lights with light-emitting diodes (LEDs) as light sources. LED lights are cold light sources, and their spectral radiation characteristics are invisible to infrared detection. Therefore, single-band detectors can no longer meet the infrared vision system's detection needs for modern airport lights, and it is necessary to increase the sensor's ability to detect visible light, so that multi-band EVS systems have become the mainstream trend of development.

[0003] However, multi-band infrared detection still faces several problems. First, the fusion of images of each band can retain the characteristic information of each band as much as possible, such as LED lighting information of visible light, runway and thermal light source information of short-wave infrared, and background information of long-wave infrared; second, the efficiency of fusing multi-band images is still relatively low. Summary of the invention

[0004] The embodiments of this specification provide a multi-band image fusion method and device based on significant features to solve the technical problem of how to improve the multi-band image fusion effect and efficiency.

[0005] To solve the above technical problems, the embodiments of this specification provide the following technical solutions:

[0006] The embodiment of this specification provides a multi-band image fusion method based on significant features, the method comprising:

[0007] For any target band image in the multi-band image group, determining a base layer image and a detail layer image of the target band image;

[0008] Obtain the significant features of each target band image in the multi-band image group, and use the significant features of each target band image to obtain the fusion coefficient matrix corresponding to the detail layer image of each target band image;

[0009] The detail layer images of each target band image are fused using the fusion coefficient matrix corresponding to the detail layer image of each target band image to obtain the detail layer fusion result; and the base layer images of each target band image are fused to obtain the base layer fusion result; and the fusion result of each target band image is obtained according to the base layer fusion result and the detail layer fusion result.

[0010] Optionally, determining the base layer image of the target band image includes:

[0011] The mean filtering algorithm is used to determine the base layer image of the target band image.

[0012] Optionally, determining the detail layer image of the target band image includes:

[0013] The detail layer image of the target band image is determined by using the target band image and the base layer image of the target band image.

[0014] Optionally, for any target band image, obtaining the significant features of the target band image includes:

[0015] According to an image obtained by a mean filtering operation on the target band image and an image obtained by a median filtering operation on the target band image, a significant feature of the target band image is obtained.

[0016] Optionally, for any target band image, obtaining the significant features of the target band image includes:

[0017] The absolute value of the result obtained by subtracting the image obtained by the median filtering operation on the target band image from the image obtained by the mean filtering operation on the target band image is used as the salient feature of the target band image.

[0018] Optionally, the detail layer images of each target band image are fused using a fusion coefficient matrix corresponding to the detail layer image of each target band image, including:

[0019] The detail layer images of each target band image are weightedly fused; wherein, for any target band image, the fusion coefficient matrix corresponding to the detail layer image of the target band image is used as the weight corresponding to the detail layer image of the target band image in the weighted fusion.

[0020] Optionally, fusing the base layer images of each target band image includes:

[0021] The weights corresponding to the base layer images of each target band image are determined, so as to perform weighted fusion on the base layer images of each target band image.

[0022] Optionally, according to the base layer fusion result and the detail layer fusion result, a fusion result of each target band image is obtained, including:

[0023] The base layer fusion result and the detail layer fusion result are fused to obtain the fusion results of each target band image.

[0024] Optionally, the multi-band image group includes a visible light image, a short-wave infrared image and a long-wave infrared image;

[0025] and / or,

[0026] Target band images include visible light images, short-wave infrared images and long-wave infrared images.

[0027] The embodiment of this specification provides a three-light fusion device based on significant features, the device comprising:

[0028] A layering module, for determining a base layer image and a detail layer image of any target band image in the multi-band image group;

[0029] The feature module is used to obtain the significant features of each target band image in the multi-band image group, and obtain the fusion coefficient matrix corresponding to the detail layer image of each target band image by using the significant features of each target band image;

[0030] The fusion module is used to fuse the detail layer images of each target band image using the fusion coefficient matrix corresponding to the detail layer image of each target band image to obtain the detail layer fusion result; and to fuse the base layer images of each target band image to obtain the base layer fusion result; and to obtain the fusion result of each target band image according to the base layer fusion result and the detail layer fusion result.

[0031] At least one of the above technical solutions adopted in the embodiments of this specification can achieve the following beneficial effects:

[0032] By performing image layering, feature extraction and layer fusion, the fusion effect and efficiency of multi-band images can be effectively improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the drawings required for use in the embodiments of this specification or the prior art description are briefly described below. Obviously, the following only describes the drawings required for use in some embodiments of this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor.

[0034] Figure 1It is a flowchart of the multi-band image fusion method based on significant features in the first embodiment of this specification.

[0035] Figure 2 It is a schematic diagram of the framework of the multi-band image fusion method based on significant features in the first embodiment of this specification.

[0036] Figure 3 This is an example diagram of a visible light image in the first embodiment of this specification.

[0037] Figure 4 This is an example of a short-wave infrared image in the first embodiment of this specification.

[0038] Figure 5 This is an example diagram of a long-wave infrared image in the first embodiment of this specification.

[0039] Figure 6 This is an example diagram of the multi-band image fusion process in the first embodiment of this specification.

[0040] Figure 7 This is an example diagram of the multi-band image fusion result in the first embodiment of this specification.

[0041] Figure 8 It is a structural diagram of a multi-band image fusion device based on significant features in the second embodiment of this specification. DETAILED DESCRIPTION

[0042] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments involved in the specific implementation methods are only part of the embodiments of this application, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments in the specific implementation methods without making creative work should belong to the scope of protection of this application.

[0043] The first embodiment of this specification (hereinafter referred to as "embodiment one") provides a multi-band image fusion method based on significant features. The execution subject of embodiment one can be a terminal (including but not limited to mobile phones, computers, pads, wearable devices, televisions) or a server or an operating system or an application, etc., that is, the execution subject can be diverse and can be set, used or changed as needed. In addition, a third-party application can also assist the execution subject in executing embodiment one. For example, the data processing method in embodiment one can be executed by a server, and an application corresponding to the server (corresponding to the server) can be installed on the terminal. Data can be transmitted between the terminal or application and the server, and data can be collected, input, output or processed through the terminal or application, thereby assisting the server in executing the method in embodiment one.

[0044] like Figure 1 and Figure 2 As shown, the multi-band image fusion method based on significant features provided in the first embodiment includes:

[0045] S101: for any target band image in a multi-band image group, determining a base layer image and a detail layer image of the target band image;

[0046] In the first embodiment, multiple images of different bands or different optical paths can be taken as a multi-band image group or a multi-optical path image group, for example, the multi-band image group can include a visible light image, a short-wave infrared image, and a long-wave infrared image; and / or, the images of each band in the multi-band image group can be collectively referred to as an image of a target band (referred to as "target band image"), for example, the target band image can include a visible light image, a short-wave infrared image, and a long-wave infrared image, so the target band in the first embodiment does not refer to a single band. Specifically, the multi-band image group can include a visible light image, a short-wave infrared image, and a long-wave infrared image, and the visible light image, the short-wave infrared image, and the long-wave infrared image can all be referred to as target band images. From the perspective of the band, short-wave infrared generally refers to a wavelength range of 0.9-1.7μm, which can also be extended to a wavelength range of 0.7-2.5μm; long-wave infrared generally refers to a wavelength range of 7.5-14μm; and visible light generally refers to a wavelength range of 380-880nm. However, these bands are not fixed.

[0047] In the first embodiment, the target band image in the multi-band image group is a target band image of the same frame, such as a visible light image, a short-wave infrared image, and a long-wave infrared image of the same frame. In practical applications, a multi-band detector (including a multi-band EVS device), a visible light detector, a short-wave infrared detector, a long-wave infrared detector, or other suitable devices can generally be used to collect the target band image.

[0048] For any target band image in the multi-band image group, a base layer image and a detail layer image of the target band image may be determined (this operation may be referred to as "image layering"). Determining the base layer image of the target band image may include: determining the base layer image of the target band image using a mean filtering algorithm. Determining the detail layer image of the target band image may include: determining the detail layer image of the target band image through the target band image and the base layer image of the target band image.

[0049] Taking the target band images including visible light images, short-wave infrared images and long-wave infrared images as an example, the visible light image can be recorded as φ1(x, y), the short-wave infrared image can be recorded as φ2(x, y), and the long-wave infrared image can be recorded as φ3(x, y). These three target band images can be used as original images or raw images and input into the execution body of the first embodiment to determine their respective base layer images and detail layer images (which can be called "three-light decomposition").

[0050] Among them, the base layer image of the target band image determined by the mean filtering algorithm can be shown as follows:

[0051]

[0052] In the formula, represents the base layer image of φ1(x,y), represents the base layer image of φ2(x,y), Represents the base layer image of φ3(x,y), μ(x,y) is a 6*6 mean filter, which is used to perform loop filtering on the image φ1(x,y) or φ2(x,y) or φ3(x,y), assign the average value of pixels in the 6*6 neighborhood to the central pixel, and traverse each pixel in the processed image.

[0053] The detail layer image of the target band image is determined by the target band image and the base layer image of the target band image, as shown in the following formula:

[0054]

[0055] In the formula, represents the detail layer image of φ1(x,y), represents the detail layer image of φ2(x,y), Represents the detail layer image of φ3(x,y).

[0056] In the first embodiment, the operations of determining the base layer image and the detail layer image of each target band image may be performed independently or in parallel.

[0057] S103: Obtaining the significant features of each target band image in the multi-band image group, and using the significant features of each target band image to obtain the fusion coefficient matrix corresponding to the detail layer image of each target band image;

[0058] In the first embodiment, the significant features of each target band image in the multi-band image group can be obtained, and the significant features refer to the high-frequency areas in the image. Wherein, for any target band image, obtaining the significant features of the target band image can include: obtaining the significant features of the target band image according to the image obtained by the mean filtering operation on the target band image and the image obtained by the median filtering operation on the target band image. Specifically, for any target band image, obtaining the significant features of the target band image can include: subtracting the image obtained by the median filtering operation on the target band image from the image obtained by the mean filtering operation on the target band image, and taking the absolute value of the result as the significant feature of the target band image.

[0059] Using the above example, the salient features of each target band image can be obtained as shown in the following formula:

[0060]

[0061] In the formula, represents the image obtained after φ1(x,y) is subjected to median filtering or operation, ξ1(x,y) represents the significant feature of φ1(x,y), represents the image obtained after φ2(x,y) is subjected to median filtering or operation, ξ2(x,y) represents the significant feature of φ2(x,y), represents the image obtained after φ3(x, y) is subjected to median filtering operation or calculation, and ξ3(x, y) represents the significant feature of φ3(x, y).

[0062] In the first embodiment, the fusion coefficient matrices corresponding to the detail layer images of each target band image can be obtained by utilizing the salient features of each target band image.

[0063] Using the above example, the above fusion coefficient matrix can be obtained as shown in the following formula:

[0064]

[0065] In the formula, ψ1(x,y) represents the fusion coefficient matrix corresponding to the detail layer image of φ1(x,y), ψ2(x,y) represents the fusion coefficient matrix corresponding to the detail layer image of φ2(x,y), and ψ3(x,y) represents the fusion coefficient matrix corresponding to the detail layer image of φ3(x,y). According to the formula, no matter how many target band images there are in the multi-band image group, the calculation method of the fusion coefficient matrix corresponding to the detail layer image of a single target band image is: divide the salient features of the detail layer image of the target band image by the sum of the salient features of the detail layer images of each target band image in the multi-band image group.

[0066] S105: using the fusion coefficient matrix corresponding to the detail layer image of each target band image, the detail layer images of each target band image are fused to obtain the detail layer fusion result; and, the base layer images of each target band image are fused to obtain the base layer fusion result; according to the base layer fusion result and the detail layer fusion result, the fusion result of each target band image is obtained.

[0067] In the first embodiment, the detail layer images of each target band image can be fused using the fusion coefficient matrix corresponding to the detail layer image of each target band image to obtain a detail layer fusion result. Among them, using the fusion coefficient matrix corresponding to the detail layer image of each target band image to fuse the detail layer images of each target band image can include: weighted fusion of the detail layer images of each target band image; and, for any target band image, using the fusion coefficient matrix corresponding to the detail layer image of the target band image as the weight corresponding to the detail layer image of the target band image in the weighted fusion (the description of the fusion coefficient matrix and the weight can be reflected together Figure 2 Specifically, weighted fusion may be to multiply the detail layer images of each target band image by their corresponding weights and then add them together.

[0068] Using the above example, the fusion of detail layer images of each target band image can be expressed as follows:

[0069]

[0070] In the formula, ψ1(x,y) is The corresponding weight, ψ2(x,y) is The corresponding weight, ψ3(x,y) is Corresponding weights are used to fuse these detail layer images using the matrix weighting strategy, i.e. weighted fusion. The detail layer fusion result is φ D (x,y).

[0071] In the first embodiment, the base layer images of each target band image can be fused to obtain a base layer fusion result. The fusion of the base layer images of each target band image can include: determining the weights corresponding to the base layer images of each target band image, thereby weighted fusion of the base layer images of each target band image. Specifically, the weighted fusion can be to multiply the base layer images of each target band image by their corresponding weights and then add them together.

[0072] Using the above example, the fusion of the base layer images of each target band image can be shown as follows:

[0073]

[0074] In the formula, α1, α2, and α3 (the sum of the three equals 1) represent The corresponding weight (i.e. weight coefficient), φ B (x,y) means The fusion result of the base layer is the fusion result of the base layer. Among them, α1, α2, and α3 can be set or changed according to the actual application scenario.

[0075] In the first embodiment, the fusion results of each target band image can be obtained according to the base layer fusion results and the detail layer fusion results, that is, the fusion results of the multi-band image group. Among them, according to the base layer fusion results and the detail layer fusion results, the fusion results of each target band image can be obtained, which can include: fusing the base layer fusion results and the detail layer fusion results (also called "double-layer fusion") to obtain the fusion results of each target band image.

[0076] In the first embodiment, the fusion of the base layer fusion result and the detail layer fusion result may be performed by adding the two.

[0077] Using the above example, the fusion of the base layer fusion result and the detail layer fusion result can be expressed as follows:

[0078] γ(x,y)=φ B (x,y)+φ D (x,y);

[0079] In the formula, γ(x,y) represents the fusion result of each target band image, that is, the fusion result of the multi-band image group.

[0080] The effect of the first embodiment is described below by an example:

[0081] Figure 3 represents a visible light image, Figure 4 represents a short-wave infrared image, Figure 5 represents a long-wave infrared image, Figures 3 to 5 Can be used as a multi-band image group.

[0082] Reference Figure 2 and Figure 6 , Figure 3 , Figure 4 , Figure 5 After three-light decomposition, the base layer (i.e. base layer image) and detail layer (i.e. detail layer image) are obtained, and the salient features (also called visual salient features) of each target band image are obtained. The weights corresponding to each detail layer image (i.e. Figure 2 "Weight Calculation" in or Figure 6 The “weight map” in ) and the fusion of detail layer images (i.e. Figure 2 or Figure 6 ) and the fusion of the base layer image (i.e. Figure 2 or Figure 6 The base layer fusion result and the detail layer fusion result are fused (i.e. Figure 2 or Figure 6 The “three-light image fusion” in the above example is used to obtain and output the fusion result, namely Figure 2 and Figure 6 The “fused image” in Figure 7 .

[0083] In the first embodiment, for the convenience of description, the term multi-band image group is used. However, in actual applications, after obtaining each target band image, the multi-band image fusion can be performed according to the method of the first embodiment, which does not mean that a grouping operation or other operations must be performed to divide each target band image into a group. That is, each target band image to be fused can be regarded as a group, but it is not necessary to form a group through a specific operation.

[0084] The examples given in the first embodiment generally take visible light images, short-wave infrared images and long-wave infrared images as examples. If the target band image is an image of other bands or also includes images of other bands, the multi-band image fusion process can be implemented with reference to the first embodiment.

[0085] The image matrix operations in the first embodiment, such as addition and subtraction operations, can specifically be pixel addition and subtraction operations, that is, the addition operation can be the addition of pixels at the same position of two or more parties, and the subtraction operation can be the subtraction of pixels at the same position of two or more parties.

[0086] Embodiment 1 can achieve the following beneficial effects:

[0087] In Example 1, from the perspective of data flow, image stratification, feature extraction, fusion of detail layer and base layer, double-layer fusion and other processes are carried out to realize the fusion of multi-band images (including but not limited to visible light images, short-wave infrared images, and long-wave infrared images). It can reduce the fusion time while retaining the significant features of each band image, has the advantages of low latency and rich features, and can effectively improve the fusion effect and efficiency of multi-band images.

[0088] Among them, through image stratification, the key features of each image can be better extracted, thereby improving the multi-band image fusion effect.

[0089] The significant feature acquisition method used in Example 1 can achieve faster, more effective and adaptive extraction of image significant features, retain more feature information of each band image, such as LED lighting information of visible light, runway and thermal light source information of short-wave infrared, and background information of long-wave infrared, and has the characteristics of being simple, efficient and time-saving, which is conducive to improving the multi-band image fusion effect and efficiency.

[0090] Embodiment 1 has broad application prospects. For example, it can be used for the development, research and application of enhanced vision equipment for civil aircraft, improve the multi-band graphics fusion effect and efficiency of enhanced vision equipment, and provide pilots with fused image data containing geographic environment information, airport runway information, and airport navigation lighting information to enhance the pilot's situational awareness of the flight environment, especially to enhance the pilot's perception and recognition ability of the aircraft situation and airport targets in night scenes and low visibility scenes.

[0091] By using a principle prototype of a multi-band enhanced vision system composed of visible light, short-wave infrared, and long-wave infrared sensors (the prototype can be manufactured as needed, as long as it meets the requirements for collecting visible light images, short-wave infrared images, and long-wave infrared images), external vision images are collected under low-visibility weather conditions, and the multi-band image fusion method described in Example 1 is verified using the collected visible light images, short-wave infrared images, and long-wave infrared images. The results show that the multi-band image fusion method described in Example 1 can quickly and efficiently fuse images of each band or each channel, and has excellent feature significance, low output delay of the fused image, and meets the design requirements (for example, the performance requirement of no more than 100ms required by DO-315B (minimum performance requirements)), which proves the feasibility of Example 1 and can be implemented in engineering.

[0092] like Figure 8 As shown, the second embodiment of this specification provides a three-light fusion device based on significant features corresponding to the method described in the first embodiment, and the device includes:

[0093] The layering module 202 is used to determine the base layer image and the detail layer image of any target band image in the multi-band image group;

[0094] The feature module 204 is used to obtain the significant features of each target band image in the multi-band image group, and obtain the fusion coefficient matrix corresponding to the detail layer image of each target band image by using the significant features of each target band image;

[0095] The fusion module 206 is used to fuse the detail layer images of each target band image using the fusion coefficient matrix corresponding to the detail layer image of each target band image to obtain a detail layer fusion result; and to fuse the base layer images of each target band image to obtain a base layer fusion result; and to obtain the fusion results of each target band image based on the base layer fusion result and the detail layer fusion result.

[0096] Optionally, determining the base layer image of the target band image includes:

[0097] The mean filtering algorithm is used to determine the base layer image of the target band image.

[0098] Optionally, determining the detail layer image of the target band image includes:

[0099] The detail layer image of the target band image is determined by using the target band image and the base layer image of the target band image.

[0100] Optionally, for any target band image, obtaining the significant features of the target band image includes:

[0101] According to an image obtained by a mean filtering operation on the target band image and an image obtained by a median filtering operation on the target band image, a significant feature of the target band image is obtained.

[0102] Optionally, for any target band image, obtaining the significant features of the target band image includes:

[0103] The absolute value of the result obtained by subtracting the image obtained by the median filtering operation on the target band image from the image obtained by the mean filtering operation on the target band image is used as the salient feature of the target band image.

[0104] Optionally, the detail layer images of each target band image are fused using a fusion coefficient matrix corresponding to the detail layer image of each target band image, including:

[0105] The detail layer images of each target band image are weightedly fused; wherein, for any target band image, the fusion coefficient matrix corresponding to the detail layer image of the target band image is used as the weight corresponding to the detail layer image of the target band image in the weighted fusion.

[0106] Optionally, fusing the base layer images of each target band image includes:

[0107] The weights corresponding to the base layer images of each target band image are determined, so as to perform weighted fusion on the base layer images of each target band image.

[0108] Optionally, according to the base layer fusion result and the detail layer fusion result, a fusion result of each target band image is obtained, including:

[0109] The base layer fusion result and the detail layer fusion result are fused to obtain the fusion results of each target band image.

[0110] Optionally, the multi-band image group includes a visible light image, a short-wave infrared image and a long-wave infrared image;

[0111] and / or,

[0112] Target band images include visible light images, short-wave infrared images and long-wave infrared images.

[0113] The above embodiments can be used in combination.

[0114] The above is only an embodiment of this specification and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A multi-band image fusion method based on salient features, characterized in that: The method comprises: For any target band image in the multi-band image group, determining a base layer image and a detail layer image of the target band image; Obtain the significant features of each target band image in the multi-band image group, and use the significant features of each target band image to obtain the fusion coefficient matrix corresponding to the detail layer image of each target band image; The detail layer images of each target band image are fused using the fusion coefficient matrix corresponding to the detail layer image of each target band image to obtain the detail layer fusion result; and the base layer images of each target band image are fused to obtain the base layer fusion result; and the fusion result of each target band image is obtained according to the base layer fusion result and the detail layer fusion result.

2. The method according to claim 1, characterized in that Determine the base layer image of the target band image includes: The mean filtering algorithm is used to determine the base layer image of the target band image.

3. The method according to claim 1 or 2, characterized in that Determine the detail layer image of the target band image includes: The detail layer image of the target band image is determined by using the target band image and the base layer image of the target band image.

4. The method according to claim 1, characterized in that For any target band image, obtain the significant features of the target band image, including: According to an image obtained by a mean filtering operation on the target band image and an image obtained by a median filtering operation on the target band image, a significant feature of the target band image is obtained.

5. The method according to claim 1 or 4, characterized in that For any target band image, obtain the significant features of the target band image, including: The absolute value of the result obtained by subtracting the image obtained by the median filtering operation on the target band image from the image obtained by the mean filtering operation on the target band image is used as the salient feature of the target band image.

6. The method according to claim 1, characterized in that The detail layer images of each target band image are fused using the fusion coefficient matrix corresponding to the detail layer images of each target band image, including: The detail layer images of each target band image are weightedly fused; wherein, for any target band image, the fusion coefficient matrix corresponding to the detail layer image of the target band image is used as the weight corresponding to the detail layer image of the target band image in the weighted fusion.

7. The method according to claim 1, characterized in that The fusion of the base layer images of each target band image includes: The weights corresponding to the base layer images of each target band image are determined, so as to perform weighted fusion on the base layer images of each target band image.

8. The method according to claim 1, characterized in that According to the base layer fusion result and the detail layer fusion result, the fusion result of each target band image is obtained, including: The base layer fusion result and the detail layer fusion result are fused to obtain the fusion results of each target band image.

9. The method according to claim 1, characterized in that The multi-band image group includes a visible light image, a short-wave infrared image and a long-wave infrared image; and / or, Target band images include visible light images, short-wave infrared images and long-wave infrared images.

10. A three-light fusion device based on significant features, characterized in that: The device comprises: A layering module, for determining a base layer image and a detail layer image of any target band image in the multi-band image group; The feature module is used to obtain the significant features of each target band image in the multi-band image group, and obtain the fusion coefficient matrix corresponding to the detail layer image of each target band image by using the significant features of each target band image; The fusion module is used to fuse the detail layer images of each target band image using the fusion coefficient matrix corresponding to the detail layer image of each target band image to obtain the detail layer fusion result; and to fuse the base layer images of each target band image to obtain the base layer fusion result; and to obtain the fusion result of each target band image according to the base layer fusion result and the detail layer fusion result.

Citation Information

Cited By

  • Early cancerization risk early warning method and device based on multi-band image fusion

    CN120183716A