Visible light and infrared image fusion method based on D-S theory weighted optimization
By employing a weighted optimization method based on DS theory, the uncertainties and computational complexity issues in visible light and infrared image fusion are resolved, generating high-quality, real-time image fusion results and improving the quality and computational efficiency of the fused images.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-31
- Publication Date
- 2026-05-15
AI Technical Summary
Existing visible light and infrared image fusion methods suffer from limited ability to handle uncertain information, low quality of fusion results, and high computational complexity, making it difficult to meet real-time requirements.
A weighted optimization method based on DS theory is adopted. By SIFT-OCT registration and denoising, edge, texture and brightness features of visible light and infrared images are extracted to generate trust functions. These functions are then fused using DS theory and weighted optimization methods. Cross-modal guided filtering and NSCT domain optimization are used to generate high-quality fused images.
It effectively handles the uncertain information of multi-source images, makes full use of complementary characteristics, generates high-quality fused images, reduces computational complexity, achieves efficient real-time fusion, preserves detailed information and highlights thermal radiation features.
Smart Images
Figure CN122049581A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of image processing technology, specifically relating to a visible light and infrared image fusion method based on DS theory weighted optimization. Background Technology
[0002] Currently, multi-source image fusion technology is widely used in fields such as security monitoring, medical imaging, autonomous driving, and target recognition. Common visible light and infrared image fusion methods include: (1) Wavelet transform-based method: decompose visible light and infrared images into different frequency bands and then fuse them.
[0003] (2) Principal Component Analysis (PCA) based method: PCA is used to extract the main components of the image and then fuse them.
[0004] (3) Weighted average method: Based on the characteristics of visible light and infrared images, different weights are assigned for fusion.
[0005] (4) Region feature-based method: Fusion is performed based on the features of the image region (such as edges, textures, etc.).
[0006] These methods can achieve the fusion of visible light and infrared images to some extent, but they generally suffer from the following problems: (1) Limited ability to process uncertain information: When the complementary information of visible light and infrared images is inconsistent or conflicting, it cannot effectively process uncertainty.
[0007] (2) The quality of the fusion result is not high: the fused image may lose detailed information or introduce noise, which affects the accuracy of subsequent target detection and recognition.
[0008] (3) High computational complexity: Some methods require complex computation processes, making it difficult to achieve efficient real-time fusion.
[0009] The main disadvantages of existing technical solutions include: (1) In the process of fusion of visible light and infrared images, the uncertainty of multi-source image information was not fully considered, resulting in insufficient reliability of the fusion result.
[0010] (2) Traditional fusion methods do not make full use of the complementary information of visible light and infrared images, and the fusion results may lack detailed information or thermal radiation characteristics.
[0011] (3) The computational complexity is high, making it difficult to meet real-time requirements. Summary of the Invention
[0012] (a) Technical problems to be solved The technical problem to be solved by this invention is: how to provide a visible light and infrared image fusion method based on DS theory weighted optimization.
[0013] (II) Technical Solution To address the aforementioned technical problems, this invention provides a visible light and infrared image fusion method based on DS theory weighted optimization, the method comprising the following steps: Step 1: Image preprocessing: Register and denoise the visible light and infrared images; Step 2: Feature Extraction: Extract key features (such as edges, texture, brightness, etc.) from visible light and infrared images; Step 3: Trust Function Generation: Based on the feature extraction results, generate trust functions for visible light and infrared images to quantify the confidence of different information sources; Step 4: Determine the weighted optimization fusion rule: Combine DS theory and weighted optimization method to fuse the trust function and generate the final fused image; Step 5: Generation and optimization of fusion results.
[0014] In step 1, the SIFT-OCT registration algorithm is used to register the visible light and infrared images at the sub-pixel level, and to perform noise reduction processing to ensure the spatial consistency of the two images and reduce noise interference.
[0015] In step 2, key features of the visible light and infrared images are extracted, including: (1) Edge features: edge information is extracted from visible light images using the Canny edge detection algorithm, and characteristic extraction is performed from infrared images using the directional gradient histogram combined with the thermal radiation intensity distribution; (2) Texture features: Texture information is extracted using the Gray-Level Co-occurrence Matrix (GLCM); (3) Brightness characteristics: Calculate the brightness value of the image to reflect the distribution of light and dark in the image.
[0016] In step 1, based on the feature extraction results, a trust function is generated for the visible light and infrared images. The trust function represents the confidence level of each information source in a specific region, and its calculation formula is as follows: Where A is the target area. is the weight value of region i, representing the importance of this region's contribution to the overall information after feature extraction. Ω represents the entire image region, where i is the index value representing a sub-region within the target region A, used to calculate the sum of the weights of these sub-regions; j is the index value representing a sub-region within the entire image region Ω, used to calculate the sum of the weights of all sub-regions, serving as the normalized denominator.
[0017] In step 4, the trust functions are fused by combining DS theory and weighted optimization methods; the fusion rules are as follows: in, For the merged trust function, and These are the trust functions for visible light and infrared images, respectively. Let B be the trust function for the visible light image of region B. Let be the trust function of the infrared image of region B, where B is any subset of the entire image region Ω, including A itself.
[0018] In step 5, the final fused image is generated based on the fused trust function, and the fusion result is optimized by using cross-modal guided filtering, using the thermal radiation information of the infrared image to guide visible light detail enhancement, and optimizing the low-frequency coefficients by taking the formaldehyde average and the high-frequency coefficients by taking the maximum absolute value in the NSCT domain, thereby generating a high-quality fused image.
[0019] (III) Beneficial Effects Compared with existing technologies, this invention proposes a weighted optimization image fusion method and apparatus based on DS theory. This method can effectively handle the uncertainty information of multi-source images, make full use of the complementary characteristics of visible light and infrared images, generate high-quality fused images, and reduce computational complexity while meeting real-time requirements.
[0020] Key points of this proposal: (1) Feature extraction and trust function generation based on DS theory: By extracting key features from visible light and infrared images, the confidence of information sources is quantified, providing a reliable basis for fusion.
[0021] (2) Weighted optimization fusion rule: Combining DS theory and weighted optimization method, the information conflict problem in multi-source image fusion is solved and the quality of fusion results is improved.
[0022] (3) Efficient calculation method: By optimizing the calculation process, the complexity of DS theory in image fusion is reduced, and efficient real-time fusion is achieved.
[0023] The beneficial effects of this invention are: (1) Information complementarity: Make full use of the complementary information of visible light and infrared images to improve the quality of fused images.
[0024] (2) Uncertainty processing: The uncertainty information of multi-source images is effectively processed by DS theory to improve the reliability of the fusion result.
[0025] (3) Computational efficiency: The calculation process of DS theory has been optimized, the computational complexity has been reduced, and efficient real-time fusion can be achieved.
[0026] (4) Robustness: It can still maintain a good fusion effect even under noise interference or poor image quality.
[0027] (5) Detail preservation: The fusion result can preserve the detail information of the visible light image while highlighting the thermal radiation characteristics of the infrared image. Attached Figure Description
[0028] Figure 1 This is a flowchart of the present invention. Detailed Implementation
[0029] To make the objectives, contents, and advantages of the present invention clearer, the specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples.
[0030] To address the aforementioned technical problems, this invention provides a visible light and infrared image fusion method based on DS theory weighted optimization, the method comprising the following steps: Step 1: Image preprocessing: Register and denoise the visible light and infrared images; Step 2: Feature Extraction: Extract key features (such as edges, texture, brightness, etc.) from visible light and infrared images; Step 3: Trust Function Generation: Based on the feature extraction results, generate trust functions for visible light and infrared images to quantify the confidence of different information sources; Step 4: Determine the weighted optimization fusion rule: Combine DS theory and weighted optimization method to fuse the trust function and generate the final fused image; Step 5: Generation and optimization of fusion results.
[0031] In step 1, the SIFT-OCT registration algorithm is used to register the visible light and infrared images at the sub-pixel level, and to perform noise reduction processing to ensure the spatial consistency of the two images and reduce noise interference.
[0032] In step 2, key features of the visible light and infrared images are extracted, including: (1) Edge features: edge information is extracted from visible light images using the Canny edge detection algorithm, and characteristic extraction is performed from infrared images using the directional gradient histogram combined with the thermal radiation intensity distribution; (2) Texture features: Texture information is extracted using the Gray-Level Co-occurrence Matrix (GLCM); (3) Brightness characteristics: Calculate the brightness value of the image to reflect the distribution of light and dark in the image.
[0033] In step 1, based on the feature extraction results, a trust function is generated for the visible light and infrared images. The trust function represents the confidence level of each information source in a specific region, and its calculation formula is as follows: Where A is the target area. is the weight value of region i, representing the importance of this region's contribution to the overall information after feature extraction. Ω represents the entire image region, where i is the index value, indicating a sub-region within the target region A, used to calculate the sum of the weights of these sub-regions. j is the index value, indicating a sub-region within the entire image region Ω, used to calculate the sum of the weights of all sub-regions, serving as the normalized denominator.
[0034] In step 4, the trust functions are fused by combining DS theory and weighted optimization methods; the fusion rules are as follows: in, For the merged trust function, and These are the trust functions for visible light and infrared images, respectively. Let B be the trust function for the visible light image of region B. Let be the trust function of the infrared image of region B, where B is any subset of the entire image region Ω (including A itself).
[0035] In step 5, the final fused image is generated based on the fused trust function, and the fusion result is optimized by using cross-modal guided filtering, using the thermal radiation information of the infrared image to guide visible light detail enhancement, and optimizing the low-frequency coefficients by taking the formaldehyde average and the high-frequency coefficients by taking the maximum absolute value in the NSCT domain, thereby generating a high-quality fused image.
[0036] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.
Claims
1. A visible light and infrared image fusion method based on DS theory weighted optimization, characterized in that, The method includes the following steps: Step 1: Image preprocessing: Register and denoise the visible light and infrared images; Step 2: Feature Extraction: Extract key features from visible light and infrared images; Step 3: Trust Function Generation: Based on the feature extraction results, generate trust functions for visible light and infrared images to quantify the confidence of different information sources; Step 4: Determine the weighted optimization fusion rule: Combine DS theory and weighted optimization method to fuse the trust function and generate the final fused image; Step 5: Generation and optimization of fusion results.
2. The visible light and infrared image fusion method based on DS theory weighted optimization as described in claim 1, characterized in that, In step 1, the SIFT-OCT registration algorithm is used to register the visible light and infrared images at the sub-pixel level, and to perform noise reduction processing to ensure the spatial consistency of the two images and reduce noise interference.
3. The visible light and infrared image fusion method based on DS theory weighted optimization as described in claim 2, characterized in that, In step 2, key features of the visible light and infrared images are extracted, including: (1) Edge features: edge information is extracted from visible light images using the Canny edge detection algorithm, and characteristic extraction is performed from infrared images using the directional gradient histogram combined with the thermal radiation intensity distribution; (2) Texture features: Texture information is extracted using the Gray-Level Co-occurrence Matrix (GLCM); (3) Brightness characteristics: Calculate the brightness value of the image to reflect the distribution of light and dark in the image.
4. The visible light and infrared image fusion method based on DS theory weighted optimization as described in claim 3, characterized in that, In step 1, a trust function is generated for the visible light and infrared images based on the feature extraction results; The trust function represents the confidence level of each information source in a specific region. The calculation formula is as follows: Where A is the target area. is the weight value of region i, representing the importance of this region's contribution to the overall information after feature extraction. It represents the entire image region, and i is the index value, which represents a sub-region in the target region A and is used to calculate the sum of the weights of these sub-regions; j is the index value, representing a sub-region within the entire image region Ω, used to calculate the sum of the weights of all sub-regions, serving as the normalized denominator.
5. The visible light and infrared image fusion method based on DS theory weighted optimization as described in claim 4, characterized in that, In step 4, the trust functions are fused by combining DS theory and weighted optimization methods; the fusion rules are as follows: in, For the merged trust function, and These are the trust functions for visible light and infrared images, respectively. Let B be the trust function for the visible light image of region B. Let be the trust function of the infrared image of region B, where B is any subset of the entire image region Ω, including A itself.
6. The visible light and infrared image fusion method based on DS theory weighted optimization as described in claim 5, characterized in that, In step 5, the final fused image is generated based on the fused trust function, and the fusion result is optimized by using cross-modal guided filtering, using the thermal radiation information of the infrared image to guide visible light detail enhancement, and optimizing the low-frequency coefficients by taking the formaldehyde average and the high-frequency coefficients by taking the maximum absolute value in the NSCT domain, thereby generating a high-quality fused image.
7. The visible light and infrared image fusion method based on DS theory weighted optimization as described in claim 5, characterized in that, The proposed method can effectively handle the uncertain information of multi-source images, make full use of the complementary characteristics of visible light and infrared images to generate high-quality fused images, and at the same time reduce computational complexity and meet real-time requirements.
8. The visible light and infrared image fusion method based on DS theory weighted optimization as described in claim 5, characterized in that, The method extracts key features from visible light and infrared images, quantifies the confidence level of information sources, and provides a reliable basis for fusion.
9. The visible light and infrared image fusion method based on DS theory weighted optimization as described in claim 5, characterized in that, The method addresses the information conflict problem in multi-source image fusion and improves the quality of the fusion results.
10. The visible light and infrared image fusion method based on DS theory weighted optimization as described in claim 5, characterized in that, The method fully utilizes the complementary information from visible light and infrared images to improve the quality of the fused image.