A medical image fusion method in FT domain
By employing a medical image fusion method in the FT domain, utilizing FT transform and GFRW model, the limitations of different medical image imaging technologies are addressed, achieving comprehensive fusion of image information and detailed extraction, thereby improving diagnostic accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-10-21
- Publication Date
- 2026-04-14
AI Technical Summary
Existing medical imaging technologies have limitations, making it difficult to provide comprehensive and accurate pathological information, which affects the diagnosis and treatment of diseases.
A medical image fusion method in the FT domain is adopted. By FT transform and L-scale decomposition, the ISML and SF values of low-frequency and high-frequency sub-band images are calculated. The GFRW model and SF model are used for image fusion. Finally, the final image is reconstructed by inverse FT transform.
While preserving the main information of the source image, image details and edge information were successfully extracted and fused to obtain ideal visual and objective evaluation results.
Smart Images
Figure CN112215922B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to medical image information processing technology, specifically to a medical image fusion method based on the GFRW model and the SF model in the FT domain. Background Technology
[0002] Currently, medical imaging has become a crucial basis for clinical diagnosis and treatment planning in the medical field. However, different imaging sensors typically have certain limitations. For example, CT images provide excellent imaging of areas such as bones, but their ability to detect soft tissues is generally limited; MRI images can provide high-resolution anatomical information for soft tissues, but their sensitivity in diagnosing areas like bones is not as high as CT images. Furthermore, functional imaging techniques such as PET and SPECT images can reflect the body's metabolic information, thus playing a vital role in the diagnosis of vascular diseases and tumor detection, but their spatial resolution is usually low. Against this backdrop, the effective fusion of images obtained from multiple different imaging sensors can undoubtedly provide doctors and patients with richer and more accurate pathological information, helping doctors make accurate judgments and improving disease cure rates. Summary of the Invention
[0003] The main objective of this invention is to provide a medical image fusion method in the FT domain, which solves the current problem of too many types of medical images and the limitations of each type.
[0004] The technical solution adopted in this invention is: a medical image fusion method in the FT domain, comprising the following steps:
[0005] Step 1: Perform FT transform on all medical source images to be fused. After L-fold scale decomposition, each source image will yield one low-frequency sub-band image and 8*L high-frequency sub-band images.
[0006] Step 2: Calculate the ISML corresponding to each pixel in the low-frequency sub-band image; obtain the initial fusion decision map; use the GF model to smooth the initial fusion decision map at three different scales to obtain the smoothed fusion decision map; further process the smoothed fusion decision map to obtain the final fusion decision map; complete the fusion of the low-frequency sub-band image.
[0007] Step 3: Process the coefficients of the high-frequency subband image; calculate the SF value of each pixel in the high-frequency subband image; complete the fusion of the high-frequency subband images;
[0008] Step 4: Perform an inverse Fourier transform on the high-frequency subband image and the low-frequency subband image of the final fused image to obtain the final fused image F.
[0009] Further, step 1 includes: inputting all medical source images to be fused, and performing FT transform on these medical source images respectively, with the scale decomposition level being L, (l,k) being the directional decomposition level at scale l, where 1≤l≤L, 1≤k≤8; after FT transform, each medical source image to be fused yields 1 low-frequency subband image and 8 high-frequency subband images respectively.
[0010] Furthermore, step 2, calculating the ISML corresponding to each pixel in the low-frequency sub-band image, includes:
[0011] The mathematical expression for calculating the ISML corresponding to each pixel in the low-frequency subband image is as follows:
[0012]
[0013] Accordingly, the ISML value corresponding to each pixel can be calculated using equation (2):
[0014]
[0015] Where N and T represent the window radius and threshold, respectively. N is a positive odd number, and T is 0; Equation (2) can be rewritten as:
[0016]
[0017] The scale in the ISML calculation process is defined as 3 levels, and the parameter N corresponding to the 3 levels is set to 5, 9, and 17 respectively; the ISML value for each pixel corresponding to different N values can be further rewritten as:
[0018] ISML nN (i,j)=ISML(I n (i,j),N),n=1,2,...Num,N=5,9,17 (4)
[0020] Where (i,j) represents the position of a pixel in the image, and Num is the number of medical source images to be fused.
[0021] Furthermore, obtaining the initial fusion decision mapping in step 2 includes:
[0022] Assuming there are two medical source images to be fused, the initial fusion decision mapping of the corresponding low-frequency sub-band image is obtained through the following mathematical expression:
[0023]
[0024] The values of parameter N are selected as 5, 9, and 17, and the value of scale parameter s is an integer in the interval [1, 3]. According to equation (5), if the ISML value of a pixel in the first image is greater than or equal to the ISML value of a pixel at the same position in the second image, then the pixel is in the initial fusion decision map. s The corresponding element in the value is set to 1, otherwise it is set to 0.
[0025] Furthermore, in step 2, the GF model is used to smooth the initial fusion decision map at three different scales, resulting in a smoothed fusion decision map including:
[0026] The GF model is used to smooth the initial fusion decision maps at three different scales to obtain the smoothed fusion decision maps:
[0027] GFmap s (i,j)=GF(I1(i,j),map s ,r s ,ε s ),s∈[1,3] (6)
[0028] Wherein, GF is the guiding filter function, I1 is a low-frequency sub-band image corresponding to the medical source image to be fused, referred to here as the guiding image; r s and ε s Representing the neighborhood size and regularization parameter respectively, this invention sets r... s ={5,11,19}, ε s ={0.01,0.001,0.0001}.
[0029] In the guided filter function GF, at a radius of r s The filter window is ω k Under the premise of guiding the image I1 and the smoothed fusion decision map GFmap s (i,j) satisfy the following local linear relationship:
[0030]
[0031] a k and b k The calculation formula is:
[0032]
[0033]
[0034]
[0035] Where, μ kand Let I1 represent the mean and variance of the guide image, respectively, and |ω| be the filter window ω. k The number of pixels within, For ω k The average value of the input image for internal filtering, and the regularization parameter ε. s Used to prevent a k The value is too large, which can lead to getting stuck in a local optimum.
[0036] Furthermore, step 2 involves further processing the smoothed fusion decision map to obtain the final fusion decision map, including:
[0037] For the smoothed fusion decision map GFmap s (i,j) undergoes further processing to obtain the final fusion decision mapping graph:
[0038]
[0039] Equation (11) is used for the smoothed fusion decision map GFmap. s (i,j) is simplified, and its corresponding values are labeled as greater than or equal to 0.75 or less than or equal to 0.25; if GFmap s If the value of (i,j) is greater than or equal to 0.75, the corresponding element values in the final fusion decision map remain unchanged and are still denoted as GFmap. s (i,j); if GFmap s When the value of (i,j) is less than or equal to 0.25, the value of the corresponding element in the final fusion decision map is set to 1-GFmap. s (i,j); If neither of the above two conditions is met, then the corresponding element value in the final fusion decision mapping graph is set to 0.
[0040] Furthermore, step 2, which involves fusing the low-frequency sub-band images, includes:
[0041] The three matrices Fmap derived from equation (11) s (i,j),s∈[1,3], to obtain the final fusion decision mapping graph Fmap. s (i,j):
[0042] Fmap(i,j)=max[Fmap1(i,j),Fmap2(i,j),Fmap3(i,j)] (12)
[0043] The low-frequency subband images are finally fused using equation (13):
[0044] F_L(i,j)=Fmap(i,j)*A_L(i,j)+(1-Fmap(i,j))*B_L(i,j) (13)
[0046] Where A_L(i,j), B_L(i,j) and F_L(i,j) represent the low-frequency sub-band image of source image A, the low-frequency sub-band image of source image B, and the low-frequency sub-band fused image, respectively.
[0047] Furthermore, step 3, processing the high-frequency subband image coefficients, includes:
[0048] The coefficients in the high-frequency subband image are converted to absolute values, i.e., abs(X_H(i,j)). (l,k) ), where X_H represents the high-frequency subband image of the medical source image X to be fused, X = A or B, (l,k) is the directional decomposition level at scale l, and l represents the l-th scale decomposition;
[0049] Step 3, calculating the SF value of each pixel in the high-frequency subband image, includes:
[0050] Calculate the SF value of each pixel in each high-frequency subband image:
[0051]
[0052] Where M and N represent the width and height of the medical source image, respectively;
[0053] Step 3, which involves fusing the high-frequency subband images, includes:
[0054] The high-frequency subband images are finally fused using equation (15):
[0055]
[0056] Furthermore, step 4 includes:
[0057] The inverse Fourier transform is used to fuse the low-frequency subband image F_L(i,j) and the high-frequency subband image F_H(i,j). (l,k) After integration, the final fused image F is obtained, and the corresponding mathematical expression is:
[0058]
[0059] Advantages of this invention:
[0060] Compared to single-type medical source images, the method of this invention can obtain a fused image that retains the main information of the medical source image to be fused, and successfully extracts the detailed and edge information of the source image and integrates it into a single image. The resulting image has a very ideal subjective visual effect and objective evaluation result.
[0061] In addition to the objectives, features, and advantages described above, the present invention has other objectives, features, and advantages. The invention will now be described in further detail with reference to the figures. Attached Figure Description
[0062] The accompanying drawings, which form part of this application, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.
[0063] Figure 1 This is a flowchart of the FT domain medical image fusion method of the present invention;
[0064] Figure 2 The images shown are medical source images to be fused used in the simulation experiment of this invention; where (a) is an MRI image and (b) is a PET image.
[0065] Figure 3 The simulation results are shown in the following figures: (a) is the simulation result of the Quadtree method; (b) is the simulation result of the DTCWTSR method; (c) is the simulation result of the MSVD method; (d) is the simulation result of the CNN method; (e) is the simulation result of the CBF method; (f) is the simulation result of the mPCNN method; (g) is the simulation result of the ASR method; and (h) is the simulation result of the method of the present invention. Detailed Implementation
[0066] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0067] refer to Figure 1 A medical image fusion method in the FT domain includes the following steps:
[0068] Step 1: Perform FT transform on all medical source images to be fused. After L-fold scale decomposition, each source image will yield one low-frequency sub-band image and 8*L high-frequency sub-band images.
[0069] Step 2: Calculate the ISML (Improved Sum of Modified Laplacian) corresponding to each pixel in the low-frequency sub-band image; obtain the initial fusion decision map; use the GF model to smooth the initial fusion decision map at three different scales to obtain the smoothed fusion decision map; further process the smoothed fusion decision map to obtain the final fusion decision map; complete the fusion of the low-frequency sub-band image;
[0070] Step 3: Process the coefficients of the high-frequency subband image; calculate the SF value of each pixel in the high-frequency subband image; complete the fusion of the high-frequency subband images;
[0071] Step 4: Perform an inverse Fourier transform on the high-frequency subband image and the low-frequency subband image of the final fused image to obtain the final fused image F.
[0072] Compared to single-type medical source images, the method of this invention can obtain a fused image that retains the main information of the medical source image to be fused, and successfully extracts the detailed and edge information of the source image and integrates it into a single image. The resulting image has a very ideal subjective visual effect and objective evaluation result.
[0073] In this embodiment, step 1 includes: inputting all medical source images to be fused, and performing FT transform on these medical source images respectively, with the scale decomposition level being L, and (l,k) being the directional decomposition level at scale l, where 1≤l≤L, 1≤k≤8; after the FT transform, each medical source image to be fused yields 1 low-frequency subband image and 8 high-frequency subband images respectively.
[0074] In this embodiment, step 2, calculating the ISML corresponding to each pixel in the low-frequency sub-band image, includes:
[0075] The mathematical expression for calculating the ISML corresponding to each pixel in the low-frequency subband image is as follows:
[0076]
[0077] Accordingly, the ISML value corresponding to each pixel can be calculated using equation (2):
[0078]
[0079] Where N and T represent the window radius and threshold, respectively. Typically, N is a positive odd number and T is 0; equation (2) can be rewritten as:
[0080]
[0081] The scale in the ISML calculation process is defined as 3 levels, and the parameter N corresponding to the 3 levels is set to 5, 9, and 17 respectively; the ISML value for each pixel corresponding to different N values can be further rewritten as:
[0082] ISML nN (i,j)=ISML(I n (i,j),N),n=1,2,...Num,N=5,9,17 (4)
[0084] Where (i,j) represents the position of a pixel in the image, and Num is the number of medical source images to be fused.
[0085] In this embodiment, obtaining the initial fusion decision mapping in step 2 includes:
[0086] Assuming there are two medical source images to be fused, the initial fusion decision mapping of the corresponding low-frequency sub-band image is obtained through the following mathematical expression:
[0087]
[0088] In this invention, the values of parameter N are selected as 5, 9, and 17, and the value range of scale parameter s is an integer within the interval [1, 3]. According to equation (5), if the ISML value of a pixel in the first image is greater than or equal to the ISML value of a pixel at the same position in the second image, then the pixel is in the initial fusion decision map map. s The corresponding element in the map is set to 1 if it is not, and 0 otherwise. Based on the three initial fusion decision maps obtained, the distribution of targets in the medical source images to be fused can be preliminarily described.
[0089] In this embodiment, step 2 uses the GF model to smooth the initial fusion decision map at three different scales, obtaining the smoothed fusion decision map as follows:
[0090] The GF model is used to smooth the initial fusion decision maps at three different scales to obtain the smoothed fusion decision maps:
[0091] GFmap s (i,j)=GF(I1(i,j),map s ,r s ,ε s ),s∈[1,3] (6)
[0092] Wherein, GF is the guiding filter function, I1 is a low-frequency sub-band image corresponding to the medical source image to be fused, referred to here as the guiding image; r s and ε sRepresenting the neighborhood size and regularization parameter respectively, this invention sets r... s ={5,11,19}, ε s ={0.01,0.001,0.0001}.
[0093] In the guided filter function GF, at a radius of r s The filter window is ω k Under the premise of guiding the image I1 and the smoothed fusion decision map GFmap s (i,j) satisfy the following local linear relationship:
[0094]
[0095] Therefore, as long as a can be calculated k and b k The numerical value of a can be used to obtain the guided filter mapping. Further, a k and b k The calculation formula is:
[0096]
[0097]
[0098]
[0099] Where, μ k and Let I1 represent the mean and variance of the guide image, respectively, and |ω| be the filter window ω. k The number of pixels within, For ω k The average value of the input image for internal filtering, and the regularization parameter ε. s Used to prevent a k The value is too large, which can lead to getting stuck in a local optimum.
[0100] In this embodiment, step 2, which further processes the smoothed fusion decision map to obtain the final fusion decision map, includes:
[0101] For the smoothed fusion decision map GFmap s (i,j) undergoes further processing to obtain the final fusion decision mapping graph:
[0102]
[0103] Equation (11) is used for the smoothed fusion decision map GFmap. s (i,j) is simplified, and its corresponding values are labeled as greater than or equal to 0.75 or less than or equal to 0.25; if GFmaps If the value of (i,j) is greater than or equal to 0.75, the corresponding element values in the final fusion decision map remain unchanged and are still denoted as GFmap. s (i,j); if GFmap s When the value of (i,j) is less than or equal to 0.25, the value of the corresponding element in the final fusion decision map is set to 1-GFmap. s (i,j); If neither of the above two conditions is met, then the corresponding element value in the final fusion decision mapping graph is set to 0.
[0104] In this embodiment, step 2, which involves fusing the low-frequency sub-band images, includes:
[0105] The three matrices Fmap derived from equation (11) s (i,j),s∈[1,3], to obtain the final fusion decision mapping graph Fmap. s (i,j):
[0106] Fmap(i,j)=max[Fmap1(i,j),Fmap2(i,j),Fmap3(i,j)] (12)
[0108] The low-frequency subband images are finally fused using equation (13):
[0109] F_L(i,j)=Fmap(i,j)*A_L(i,j)+(1-Fmap(i,j))*B_L(i,j) (13)
[0111] Where A_L(i,j), B_L(i,j) and F_L(i,j) represent the low-frequency sub-band image of source image A, the low-frequency sub-band image of source image B, and the low-frequency sub-band fused image, respectively.
[0112] In this embodiment, the processing of the high-frequency subband image coefficients in step 3 includes:
[0113] Unlike low-frequency subband images, the coefficients in high-frequency subband images obtained after FT transform contain both positive and negative numbers, and coefficients with larger absolute values often contain important details from the medical source image. Therefore, this invention performs absolute value processing on the coefficients in the high-frequency subband image, i.e., abs(X_H(i,j)). (l,k) ), where X_H represents the high-frequency subband image of the medical source image X to be fused, X = A or B, (l,k) is the directional decomposition level at scale l, and l represents the l-th scale decomposition;
[0114] Step 3, calculating the SF value of each pixel in the high-frequency subband image, includes:
[0115] Calculate the SF value of each pixel in each high-frequency subband image:
[0116]
[0117] Where M and N represent the width and height of the medical source image, respectively;
[0118] Step 3, which involves fusing the high-frequency subband images, includes:
[0119] The high-frequency subband images are finally fused using equation (15):
[0120]
[0121] In this embodiment, step 4 includes:
[0122] The inverse Fourier transform is used to fuse the low-frequency subband image F_L(i,j) and the high-frequency subband image F_H(i,j). (l,k) After integration, the final fused image F is obtained, and the corresponding mathematical expression is:
[0123]
[0124] The simulation platform for the method of this invention is a personal PC configured with an Intel(R) Core(TM) i5-4250U CPU 1.90GHz, 4GB of memory, and Matlab 2014b as the simulation software. To better understand the technical solution of this invention, this embodiment uses two medical source images (MRI image + PET image) for fusion. (Reference) Figure 1 In the figure, the two medical source images are denoted as A and B respectively, and the final fused image is denoted as F; following the technical solution of the present invention.
[0125] Simulation comparison experiment:
[0126] To verify the effectiveness of the method of the present invention, a series of simulation experiments are conducted to demonstrate that, compared with various existing conventional image fusion methods, the method of the present invention has better rationality and effectiveness:
[0127] Following the technical solution of this invention, a set of medical source images is fused, the set of source images including an MRI image (see...). Figure 2 (a) and a PET image (see Figure 2(b) of this paper describes the fusion process and compares the fusion results with several representative methods. First, the two medical source images to be fused are subjected to Fourier Transform (FT). Then, the low-frequency subband image and the high-frequency subband image are fused using the GFRW model and SF model proposed in this invention, respectively. Finally, the high-frequency subband image and the low-frequency subband image of the final fused image are subjected to inverse Fourier Transform (FT) to obtain the final fused image. Simultaneously, several representative methods, including Quadtree, DTCWTSR, MSVD, CNN, CBF, mPCNN, and ASR, are selected and compared with the corresponding methods of this invention.
[0128] Figure 3 Simulation results for eight methods are presented, demonstrating that the fusion method of this invention has excellent fusion performance and can effectively extract and fuse subject and detail information from the medical source images to be fused. Furthermore, mutual information (MI), structural similarity index (SSIM), and edge information retention (Q) were also selected as parameters. AB / F The Sum of the Correlations of Differences (SCD) and the objective quality evaluation index of the eight methods were used. Table 1 shows the objective evaluation results of the final fused images corresponding to the eight image fusion methods in the simulation experiment. It should be noted that the degree of excellence of each result in the four objective evaluation index results was ranked, and the ranking result was recorded as the superscript of each result, with the smaller the value, the better the result. In order to objectively and fairly evaluate the eight methods, this invention accumulated the ranking of the objective evaluation results of all methods, with the smaller the value, the better the performance. Among them, the ranking accumulation value of the method of this invention is 14, which is better than the other seven comparison methods.
[0129] Table 1 Objective evaluation results of eight image fusion methods
[0130]
[0131] This invention employs Fourier Transform (FT) to decompose the medical source images to be fused into multi-scale, multi-directional images. Compared to classic transforms such as wavelet transform, ridge transform, non-subsampled contour wave transform, and non-subsampled shear wave transform, this transform offers better information capture capabilities and lower computational complexity, exhibiting highly desirable performance. Furthermore, the invention combines the GFRW and SF models to perform low-frequency sub-band image fusion and high-frequency sub-band image fusion respectively. Finally, the inverse FT transform is used to reconstruct the final fused image. Simulation experiments demonstrate that the fusion method of this invention has excellent fusion performance, effectively extracting and fusing both subject and detail information from the medical source images to be fused.
[0132] The method of this invention, based on the imaging principle of multimodal medical images, comprehensively utilizes the GFRW model and SF model to effectively fuse images such as CT images, MRI images, PET images and SPECT images. It makes full use of the advantages of different image sensors and effectively fuses medical image information from different sources, thus facilitating a reasonable solution to the problem of multimodal medical image fusion. It has both high academic value and very broad practical prospects.
[0133] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A medical image fusion method in the FT domain, characterized in that, Includes the following steps: Step 1: Perform FT transform on all medical source images to be fused. After L-fold scale decomposition, each source image will yield one low-frequency sub-band image and 8*L high-frequency sub-band images. Step 2: Calculate the ISML corresponding to each pixel in the low-frequency sub-band image; obtain the initial fusion decision map; use the GF model to smooth the initial fusion decision map at three different scales to obtain the smoothed fusion decision map; further process the smoothed fusion decision map to obtain the final fusion decision map; complete the fusion of the low-frequency sub-band image. Step 3: Process the coefficients of the high-frequency subband image; calculate the SF value of each pixel in the high-frequency subband image; complete the fusion of the high-frequency subband images; Step 4: Perform an inverse Fourier transform on the high-frequency subband image and the low-frequency subband image of the final fused image to obtain the final fused image F; In step 2, the GF model is used to map the initial fusion decision at three different scales. The graph is smoothed to obtain a smoothed fusion decision mapping graph, including: The GF model is used to smooth the initial fusion decision maps at three different scales to obtain the smoothed fusion decision maps: (6) Wherein, GF is the guiding filter function, I1 is a low-frequency sub-band image corresponding to the medical source image to be fused, referred to here as the guiding image; r s and ε s Let r represent the neighborhood size and the regularization parameter, respectively. s ={5, 11, 19}, ε s ={0.01,0.001, 0.0001}; In the guided filter function GF, at a radius of r s The filter window is ω k Under the premise of guiding the image I1 and the smoothed fusion decision map GFmap s (i, j) satisfy the following local linear relationship: (7) a k and b k The calculation formula is: (8) (9) (10) Where, μ k and Let I1 represent the mean and variance of the guide image, respectively, and |ω| be the filter window ω. k The number of pixels within, For ω k The average value of the input image for internal filtering, and the regularization parameter ε. s Used to prevent a k The value is too large, which can lead to getting stuck in a local optimum.
2. The medical image fusion method in the FT domain according to claim 1, characterized in that, Place Step 1 includes: inputting all medical source images to be fused, and performing FT transform on these medical source images respectively, with the scale decomposition level being L, and (l, k) being the directional decomposition level at scale l, where 1≤l≤L, 1≤k≤8; after FT transform, each medical source image to be fused yields 1 low-frequency subband image and 8 high-frequency subband images respectively.
3. The medical image fusion method in the FT domain according to claim 1, characterized in that, Place Step 2, calculating the ISML corresponding to each pixel in the low-frequency subband image, includes: The mathematical expression for calculating the ISML corresponding to each pixel in the low-frequency subband image is as follows: (1) Accordingly, the ISML value corresponding to each pixel can be calculated using equation (2): (2) Where N and T represent the window radius and threshold, respectively; N is a positive odd number and T is 0; Equation (2) can be rewritten as: (3) The scale in the ISML calculation process is defined as 3 levels, and the parameter N corresponding to the 3 levels is set to 5, 9, and 17 respectively; the ISML value for each pixel corresponding to different N values can be further rewritten as: (4) Where (i, j) represents the position of a pixel in the image, and Num is the number of medical source images to be fused.
4. The medical image fusion method in the FT domain according to claim 1, characterized in that, Place The initial fusion decision mapping obtained in step 2 includes: Assuming there are two medical source images to be fused, the initial fusion decision mapping of the corresponding low-frequency sub-band image is obtained through the following mathematical expression: (5) The values of parameter N are selected as 5, 9, and 17, and the value range of scale parameter s is an integer within the interval [1, 3]. According to equation (5), if the ISML value of a pixel in the first image is greater than or equal to the ISML value of a pixel at the same position in the second image, then the pixel is included in the initial fusion decision map. s The corresponding element in the value is set to 1, otherwise it is set to 0.
5. The medical image fusion method in the FT domain according to claim 1, characterized in that, Place Step 2 further processes the smoothed fusion decision map to obtain the final fusion decision map, including: For the smoothed fusion decision map GFmap s (i, j) are further processed to obtain the final fusion decision mapping graph: (11) Equation (11) is used for the smoothed fusion decision map GFmap. s (i, j) is simplified, and its corresponding values are labeled as greater than or equal to 0.75 or less than or equal to 0.25; if GFmap s If the value of (i, j) is greater than or equal to 0.75, the value of the corresponding element in the final fusion decision map remains unchanged and is still denoted as GFmap. s (i, j); if GFmap s When the value of (i, j) is less than or equal to 0.25, the value of the corresponding element in the final fusion decision map is set to 1-GFmap. s (i, j); If neither of the above two conditions is met, the corresponding element value in the final fusion decision mapping graph is set to 0.
6. The medical image fusion method in the FT domain according to claim 5, characterized in that, Step 2, which involves fusing the low-frequency sub-band images, includes: The three matrices Fmap derived from equation (11) s (i, j), s [1, 3], to obtain the final fusion decision mapping graph Fmap. s (i,j): (12) The low-frequency subband images are finally fused using equation (13): (13) Where A_L(i, j), B_L(i, j) and F_L(i, j) represent the low-frequency sub-band image of source image A, the low-frequency sub-band image of source image B, and the fused low-frequency sub-band image, respectively.
7. The medical image fusion method in the FT domain according to claim 1, characterized in that, Place Step 3, which involves processing the high-frequency subband image coefficients, includes: The coefficients in the high-frequency subband image are converted to absolute values, i.e., abs(X_H(i, j)). (l, k) ), where X_H represents the high-frequency subband image of the medical source image X to be fused, X=A or B, (l, k) is the directional decomposition level at scale l, and l represents the l-th scale decomposition; Step 3, calculating the SF value of each pixel in the high-frequency subband image, includes: Calculate the SF value of each pixel in each high-frequency subband image: (14) Where M and N represent the width and height of the medical source image, respectively; Step 3, which involves fusing the high-frequency subband images, includes: The high-frequency subband images are finally fused using equation (15): (15)。 8. The medical image fusion method in the FT domain according to claim 6 or 7, characterized in that, Step 4 includes: The inverse Fourier transform is used to fuse the low-frequency subband image F_L(i, j) and the high-frequency subband image F_H(i, j). (l, k) After integration, the final fused image F is obtained, and the corresponding mathematical expression is: (16)。