CN116152071A - A Fusion and Reconstruction Method for Multi-View Anisotropic 3D Images - Google Patents

Abstract

The invention relates to a method for fusion and reconstruction of multi-view anisotropic three-dimensional images, comprising: acquiring three orthogonal MRI images of small mammalian organs; making the spatial proportions of the three orthogonal MRI images consistent to obtain image I1 , I2 and I3; extract the feature points on the reference image R1; use the corresponding relationship between the obtained reference image and the two sets of feature point pairs corresponding to the floating image to calculate the affine transformation matrix of the reference image and the floating image; get two Generate images R2 and R3; obtain three images R1', R2' and R3' after brightness adjustment; obtain a three-dimensional isotropic image, which is the final output image I _o . The present invention accurately stitches and aligns three images of different angles of view, and through the fusion and reconstruction of multiple orthogonal low-resolution images, the clarity of the original low-resolution images is greatly improved. Animal organ function research provides high-definition image data support.

Description

A Fusion and Reconstruction Method for Multi-View Anisotropic 3D Images

technical field

The invention relates to the technical field of 3D biomedical image processing, in particular to a method for fusion and reconstruction of multi-view anisotropic three-dimensional images.

Background technique

Magnetic Resonance Imaging MRI is a non-invasive imaging technique that produces three-dimensional detailed images of anatomy, often used for disease detection, diagnosis and treatment monitoring. Limited to imaging technology, current MRI imaging is mainly used for imaging human organs. The interlayer resolution that can be achieved by general medical equipment is at the millimeter level. To improve the interlayer resolution, a high-intensity imaging magnetic field is required.

At present, MRI imaging for small mammals is still very difficult. For example, the brain size of tree shrews is less than two centimeters. High-resolution imaging equipment is required to obtain images that can be used for analysis. The imaging conditions are harsh and there are few image samples. At this stage, the methods for improving the interlayer resolution of images in the later stage often require preset prior conditions and a large amount of high-quality data for learning, which is contrary to the reality that high-quality image data is difficult to obtain under actual conditions.

Contents of the invention

In order to solve the defect that the existing imaging equipment is difficult to obtain images with high interlayer resolution for MRI imaging of the tiny organs of small animals, the purpose of the present invention is to provide a method to obtain approximately orthogonal low-layer interlayer images through three scans with different viewing angles. The high-resolution images, after a series of processing, are fused to reconstruct isotropic interlayer high-resolution isotropic three-dimensional images, which provide high-definition image data support for the study of organ functions of small mammals, with high operability , a fusion reconstruction method for multi-view anisotropic 3D images with good fusion reconstruction effect.

In order to achieve the above object, the present invention adopts the following technical solution: a method for fusion and reconstruction of multi-view anisotropic three-dimensional images, the method includes the following sequential steps:

(1) Obtain three orthogonal MRI images of small mammalian organs;

(2) Convert the three orthogonal MRI images from the nii format to the v3draw format. By calculating the proportional relationship between the intra-slice resolution and the inter-slice resolution, the inter-layer pixels are interpolated using linear interpolation to make the three orthogonal MRI images The spatial proportions of the MRI images are consistent, and images I1, I2 and I3 are obtained;

(3) After linear interpolation, select an image I1 with the best quality among the three images I1, I2 and I3 as the reference image R1, and the other two images I2 and I3 as floating images, respectively in the three images R1, I2 Select a small number of feature points representing the direction of the image on and I3, and use the 2.5DHarris corner detection algorithm to extract feature points on the reference image R1;

(4) Use the CLM coherent marker point mapping matching algorithm to search for feature points that match the feature points on the reference image R1 on the two floating images I2 and I3, and use the two sets of feature point pairs corresponding to the reference image and the floating image The correspondence between calculates the affine transformation matrix of the reference image and the floating image;

(5) Carry out affine transformation to the two floating images using the affine transformation matrix, so that the two floating images I2 and I3 are completely aligned with the reference image R1, and two generated images R2 and R3 are obtained;

(6) Taking the brightness distribution of the reference image R1 as a reference, adjust the brightness of the two generated images R2 and R3 to keep the brightness distribution of the three consistent, and obtain the three images R1', R2' and R3 after brightness adjustment ';

(7) For the three images R1', R2' and R3', according to the three-dimensional gradient enhancement fusion method, the three images R1', R2' and R3' are fused and reconstructed to obtain a three-dimensional isotropic picture that is the final output Image I _o .

Described step (2) specifically comprises the following steps:

(2a) Convert the image of the anisotropic three-dimensional nii format into the v3draw format;

(2b) Calculate the upsampling multiple between image slices, and according to the proportional relationship between the internal resolution of the image and the resolution between slices, the internal resolution of the acquired image is n*n microns, and the resolution between slices is m microns, m>n, calculate the inter-chip upsampling multiple as α=m/n;

(2c) Calculate the upsampled image size according to the upsampling multiple: the initial anisotropic image size is X*Y*Z, then the interpolated output image size is X*Y*(Z*α);

(2d) Interlayer upsampling is performed on the 3D anisotropic image according to the size of the output image, the interpolation method is bilinear interpolation, and the three images are interpolated in the same way to obtain three images I1, I2 and I3.

Described step (3) specifically comprises the following steps:

(3a) Compare the three images after linear interpolation, select the image I1 with the best quality as the reference image, and name it as R1, and the two floating images are I2 and I3; the best quality refers to the largest field of view , the brightness distribution is uniform;

(3b) On the three images R1, I2 and I3, click and select 8 feature points that can represent the general direction distribution of the three images;

(3c) Using the 2.5DHarris corner detection algorithm, set the non-maximum suppression window respectively, wherein the radii of the three-dimensional non-maximum suppression window and the two-dimensional non-maximum suppression window are set to 25 and 20 respectively, in the reference image R1 Extract 700 feature points;

(3d) Filter the generated 700 feature points, and finally keep 500 feature points.

Described step (4) specifically comprises the following steps:

(4a) Using the selected feature points, calculate the rough-precision three-dimensional affine transformation matrix from the floating image I2 to the reference image R1, and from the floating image I3 to the reference image R1, and use the rough-precision three-dimensional affine transformation matrix to transform the two floating images I2 and I3 do preliminary affine transformation to the reference image R1 to obtain floating image I2' and floating image I3', and the basic directions of the three images R1, I2' and I3' are consistent;

(4b) Input the reference image R1 and use the 2.5DHarris corner detection algorithm to extract the feature points on the reference image R1, as well as the floating images I2' and I3' after orientation adjustment, and use the CLM coherent marker point mapping matching algorithm to set the search radius is 10, find the feature points matching the feature points on the reference image R1 on the floating images I2' and floating I3' by means of iterative search;

(4c) Use the inverse matrix of the coarse-precision three-dimensional affine transformation matrix to inversely transform the feature points on the floating images I2' and I3' to the floating images I2 and I3, and obtain the accurate matching of the floating images I2 and I3 with the reference image R1 Feature points, using these two sets of feature point pairs that are accurately matched, calculate the affine transformation matrix for transforming the floating image I2 to the reference image R1, and the floating image I3 to the reference image R1.

Described step (5) specifically comprises the following steps:

(5a) Generate two full-space images with the same size as the reference image R1 with a pixel value of 0, obtain the pixel value of each pixel of the two full-space images, and obtain two full-space images according to the affine transformation matrix The position of each pixel in the image in the original images of the two floating images I2 and I3;

(5b) Use the pixel values of the eight neighborhood positions around this position in the original image to perform linear interpolation on the pixel values of the generated image; if the position of the target image pixel point in the original image exceeds the boundary, then the pixel value beyond the boundary is taken as Referring to the pixel value of the position of the image R1, the generated image R2 is obtained according to the floating image I2, and the generated image R3 is obtained according to the floating image I3.

Described step (6) specifically comprises the following steps:

(6a) Select the radius of the sliding one-dimensional window as 5, and perform mean smoothing filtering on the three viewing angles of X, Y, and Z of the three images R1, R2, and R3 respectively, and obtain three smoothed and filtered images S1, S2, and S3 ;

(6b) Traversing the three images R1, R2, R3 pixel by pixel, adjusting the brightness values of the generated images R2 and R3, the pixel value R2(x, y, z) at the corresponding position after brightness matching = R2(x, y, z)*(S1(x,y,z)/S2(x,y,z)), R3(x,y,z)=R3(x,y,z)*(S1(x,y,z) /S3(x, y, z)) to obtain three images R1', R2' and R3' whose brightness has been adjusted and kept consistent.

Described step (7) specifically comprises the following steps:

(7a) Generate an all-empty image with a pixel value of 0 of the same size as the three images R1', R2' and R3';

(7b) For the empty image in step (7a), each pixel position is traversed, and for each pixel position (x, y, z) in the final output image _Io , three images R1', R2' and R3' corresponds to the position and the pixel value of the neighborhood R'(x,y,z), R'(x+1,y,z), R'(x-1,y,z), R'(x,y +1,z), R'(x,y-1,z), R'(x,y,z+1), R'(x,y,z-1);

其中，取a＝4b，b的取值范围为0至1，根据融合效果进行调整，在上式中若I_o(x,y,z)<0或者I_o(x,y,z)>255，则取

逐次遍历，得到最终输出图像I_o。(7c) According to the fusion method of three-dimensional gradient enhancement, the three images are fused in the following way: for the assignment of the pixel value of each pixel position in the final output image _Io , take

Among them, take a=4b, the value range of b is 0 to 1, and adjust according to the fusion effect. In the above formula, if I _o (x, y, z)<0 or I _o (x, y, z)> 255, then take

Traverse successively to get the final output image I _o .

It can be seen from the above technical solution that the beneficial effects of the present invention are as follows: first, the three images of different viewing angles are accurately spliced and aligned, and the images of different viewing ranges obtained from multiple viewing angles can be complemented, and the images contained in the three images Second, through the fusion and reconstruction of multiple orthogonal low-resolution images, the clarity of the original low-resolution images is greatly improved, which provides a great opportunity for the study of organ functions in small mammals. High-definition image data support; third, the present invention proposes a new image acquisition method, which can be fused and reconstructed by relying on a small number of images acquired from different perspectives, with high operability and good fusion and reconstruction effects.

Description of drawings

Fig. 1 is method flowchart of the present invention;

Figure 2 shows three interlayer low-resolution anisotropic images collected;

Fig. 3 is the result figure of interlayer interpolation of the anisotropic image of low resolution;

Fig. 4 is an effect diagram of accurate registration of two floating images to the reference image;

Fig. 5 is the effect diagram of brightness matching of three images;

Figure 6 is an effect diagram of three images after fusion and reconstruction.

Detailed ways

As shown in Figure 1, a method for fusion and reconstruction of multi-view anisotropic three-dimensional images, the method includes the steps in the following order:

(1) Obtain three orthogonal MRI images of small mammalian organs; due to the limitation of imaging technology, the inter-layer resolution of the obtained images is lower than the intra-layer resolution, and it is difficult to achieve a strict orthogonal effect during image manipulation;

(2) Convert the three orthogonal MRI images from the nii format to the v3draw format. By calculating the proportional relationship between the intra-slice resolution and the inter-slice resolution, the inter-layer pixels are interpolated using linear interpolation to make the three orthogonal MRI images The spatial proportions of the MRI images are consistent, and images I1, I2 and I3 are obtained;

(3) After linear interpolation, select an image I1 with the best quality among the three images I1, I2 and I3 as the reference image R1, and the other two images I2 and I3 as floating images, respectively in the three images R1, I2 Select a small number of feature points representing the direction of the image on and I3, and use the 2.5DHarris corner detection algorithm to extract feature points on the reference image R1;

(4) Use the CLM coherent marker point mapping matching algorithm to search for feature points that match the feature points on the reference image R1 on the two floating images I2 and I3, and use the two sets of feature point pairs corresponding to the reference image and the floating image The corresponding relationship between calculates the affine transformation matrix of the reference image and the floating image; the CLM coherent marker point mapping matching algorithm comes from the paper: Cross-Modality Coherent Registration of Whole Mouse Brains;

(5) Carry out affine transformation to the two floating images using the affine transformation matrix, so that the two floating images I2 and I3 are completely aligned with the reference image R1, and two generated images R2 and R3 are obtained;

(6) Taking the brightness distribution of the reference image R1 as a reference, adjust the brightness of the two generated images R2 and R3 to keep the brightness distribution of the three consistent, and obtain the three images R1', R2' and R3 after brightness adjustment ';

(7) For the three images R1', R2' and R3', according to the three-dimensional gradient enhancement fusion method, the three images R1', R2' and R3' are fused and reconstructed to obtain a three-dimensional isotropic picture that is the final output Image I _o .

Described step (2) specifically comprises the following steps:

(2a) Convert the image of the anisotropic three-dimensional nii format into the v3draw format;

(2b) Calculate the upsampling multiple between image slices, and according to the proportional relationship between the internal resolution of the image and the resolution between slices, the internal resolution of the acquired image is n*n microns, and the resolution between slices is m microns, m>n, calculate the inter-chip upsampling multiple as α=m/n;

(2c) Calculate the upsampled image size according to the upsampling multiple: the initial anisotropic image size is X*Y*Z, then the interpolated output image size is X*Y*(Z*α);

(2d) Interlayer upsampling is performed on the 3D anisotropic image according to the size of the output image, the interpolation method is bilinear interpolation, and the three images are interpolated in the same way to obtain three images I1, I2 and I3.

Described step (3) specifically comprises the following steps:

(3a) Compare the three images after linear interpolation, select the image I1 with the best quality as the reference image, and name it as R1, and the two floating images are I2 and I3; the best quality refers to the largest field of view , the brightness distribution is uniform;

(3b) On the three images R1, I2 and I3, click and select 8 feature points that can represent the general direction distribution of the three images;

(3c) Using the 2.5DHarris corner detection algorithm, set the non-maximum suppression window respectively, wherein the radii of the three-dimensional non-maximum suppression window and the two-dimensional non-maximum suppression window are set to 25 and 20 respectively, in the reference image R1 Extract 700 feature points;

(3d) Filter the generated 700 feature points, and finally keep 500 feature points.

Described step (4) specifically comprises the following steps:

(4a) Using the selected feature points, calculate the rough-precision three-dimensional affine transformation matrix from the floating image I2 to the reference image R1, and from the floating image I3 to the reference image R1, and use the rough-precision three-dimensional affine transformation matrix to transform the two floating images I2 and I3 do preliminary affine transformation to the reference image R1 to obtain floating image I2' and floating image I3', and the basic directions of the three images R1, I2' and I3' are consistent;

(4b) Input the reference image R1 and use the 2.5DHarris corner detection algorithm to extract the feature points on the reference image R1, as well as the floating images I2' and I3' after orientation adjustment, and use the CLM coherent marker point mapping matching algorithm to set the search radius is 10, find the feature points matching the feature points on the reference image R1 on the floating images I2' and floating I3' by means of iterative search;

(4c) Use the inverse matrix of the coarse-precision three-dimensional affine transformation matrix to inversely transform the feature points on the floating images I2' and I3' to the floating images I2 and I3, and obtain the accurate matching of the floating images I2 and I3 with the reference image R1 Feature points, using these two sets of feature point pairs that are accurately matched, calculate the affine transformation matrix for transforming the floating image I2 to the reference image R1, and the floating image I3 to the reference image R1.

Described step (5) specifically comprises the following steps:

(5a) Generate two full-space images with the same size as the reference image R1 with a pixel value of 0, obtain the pixel value of each pixel of the two full-space images, and obtain two full-space images according to the affine transformation matrix The position of each pixel in the image in the original images of the two floating images I2 and I3;

(5b) Use the pixel values of the eight neighborhood positions around this position in the original image to perform linear interpolation on the pixel values of the generated image; if the position of the target image pixel point in the original image exceeds the boundary, then the pixel value beyond the boundary is taken as Referring to the pixel value of the position of the image R1, the generated image R2 is obtained according to the floating image I2, and the generated image R3 is obtained according to the floating image I3.

Described step (6) specifically comprises the following steps:

(6a) Select the radius of the sliding one-dimensional window as 5, and perform mean smoothing filtering on the three viewing angles of X, Y, and Z of the three images R1, R2, and R3 respectively, and obtain three smoothed and filtered images S1, S2, and S3 ;

(6b) Traversing the three images R1, R2, R3 pixel by pixel, adjusting the brightness values of the generated images R2 and R3, the pixel value R2(x, y, z) at the corresponding position after brightness matching = R2(x, y, z)*(S1(x,y,z)/S2(x,y,z)), R3(x,y,z)=R3(x,y,z)*(S1(x,y,z) /S3(x, y, z)) to obtain three images R1', R2' and R3' whose brightness has been adjusted and kept consistent.

Described step (7) specifically comprises the following steps:

(7a) Generate an all-empty image with a pixel value of 0 of the same size as the three images R1', R2' and R3';

(7b) For the empty image in step (7a), each pixel position is traversed, and for each pixel position (x, y, z) in the final output image _Io , three images R1', R2' and R3' corresponds to the position and the pixel value of the neighborhood R'(x,y,z), R'(x+1,y,z), R'(x-1,y,z), R'(x,y +1,z), R'(x,y-1,z), R'(x,y,z+1), R'(x,y,z-1);

其中，取a＝4b，b的取值范围为0至1，根据融合效果进行调整，在上式中若I_o(x,y,z)<0或者I_o(x,y,z)>255，则取

逐次遍历，得到最终输出图像I_o。(7c) According to the fusion method of three-dimensional gradient enhancement, the three images are fused in the following way: for the assignment of the pixel value of each pixel position in the final output image _Io , take

Among them, take a=4b, the value range of b is 0 to 1, and adjust according to the fusion effect. In the above formula, if I _o (x, y, z)<0 or I _o (x, y, z)> 255, then take

Traverse successively to get the final output image I _o .

As shown in Figure 2, the images are three approximately orthogonal tree shrew brain MRI images acquired under a high-intensity magnetic field, and the resolution can reach the micron level. It can be seen from the images that the resolution of each image in the slice is The resolution is much higher than the resolution between image slices, and the 3D view shows that the images are obviously inconsistent in spatial scale.

As shown in Figure 3, the upsampling ratio of the three images is calculated, and the interlayer pixels are filled by bilinear interpolation. The images in the other two views are of lower quality; the image in the middle has the largest imaging coverage area and better image quality, and is used as a reference image, and the other two images are used as floating images.

As shown in Figure 4, for the subsequent image fusion, the two floating images are accurately registered and transformed to the reference image. After the transformation, the feature positions of the three images are completely aligned; at this time, the brightness distribution of the three images is not consistent.

As shown in Figure 5, in order to ensure that the brightness distribution of the three images is consistent, the brightness distribution of the reference image is used as a benchmark to perform brightness matching on the other two images, and the brightness distribution of the three images after brightness adjustment can be kept consistent .

As shown in Figure 6, for the three images that have been fully aligned and adjusted for brightness, only one view of the image is still of good quality, and the quality of the other two views is still low due to interpolation. After the fusion method of three-dimensional gradient enhancement, the results shown in Figure 6 are obtained. From the results, it can be seen that the image can guarantee the optimal effect in the three views.

In summary, the present invention accurately stitches and aligns three images of different viewing angles, can complement images of different viewing ranges acquired from multiple viewing angles, and integrate the effective information contained in the three images; The fusion reconstruction of orthogonal low-resolution images greatly improves the clarity of the original low-resolution images, and provides high-definition image data support for the study of organ functions in small mammals.

Claims (7)

1. A method for fusion reconstruction of multi-view anisotropic three-dimensional images, characterized in that the method comprises the following steps in order: (1) Acquire three orthogonal MRI images of small mammal organs; (2) The three orthogonal MRI images are converted from nii format to v3draw format. By calculating the ratio between the intra-slice resolution and the inter-slice resolution, the inter-slice pixels are interpolated using linear interpolation to make the spatial ratios of the three orthogonal MRI images consistent, thus obtaining images I1, I2, and I3. (3) After linear interpolation, the image I1 with the best quality among the three images I1, I2 and I3 is selected as the reference image R1, and the other two images I2 and I3 are used as floating images. A small number of feature points representing the image direction are selected on the three images R1, I2 and I3 respectively, and the feature points on the reference image R1 are extracted using the 2.5D Harris corner detection algorithm; (4) using the CLM coherent marker mapping matching algorithm to search for feature points on the two floating images I2 and I3 that match the feature points on the reference image R1, and using the correspondence between the two sets of feature point pairs corresponding to the reference image and the floating image to calculate the affine transformation matrix of the reference image and the floating image; (5) Using the affine transformation matrix, the two floating images are affine transformed so that the two floating images I2 and I3 are completely aligned with the reference image R1, thereby obtaining two generated images R2 and R3; (6) Taking the brightness distribution of the reference image R1 as a reference, the brightness of the two generated images R2 and R3 is adjusted to make the brightness distribution of the three images consistent, thereby obtaining three images R1', R2' and R3' after brightness adjustment; (7) For the three images R1', R2' and R3', the three images R1', R2' and R3' are fused and reconstructed according to the 3D gradient enhancement fusion method to obtain a 3D isotropic image, namely the final output image _Io . 2. The method for fusion reconstruction of multi-view anisotropic three-dimensional images according to claim 1, characterized in that: the step (2) specifically comprises the following steps: (2a) Convert the anisotropic three-dimensional nii format image into v3draw format; (2b) Calculate the upsampling multiple between image slices. According to the ratio between the intra-slice resolution and the inter-slice resolution, the intra-slice resolution of the acquired image is n*n microns, and the inter-slice resolution is m microns. m>n, and the inter-slice upsampling multiple is calculated to be α=m/n. (2c) Calculate the size of the upsampled image according to the upsampling multiple: the initial anisotropic image size is X*Y*Z, and the output image size after interpolation is X*Y*(Z*α); (2d) The three-dimensional anisotropic image is upsampled between layers according to the output image size. The interpolation method is bilinear interpolation. The three images are interpolated in the same way to obtain three images I1, I2 and I3 with the same spatial scale. 3. The method for fusion reconstruction of multi-view anisotropic three-dimensional images according to claim 1, characterized in that: the step (3) specifically comprises the following steps: (3a) Compare the three images after linear interpolation, select the image I1 with the best quality as the reference image, named R1, and the two floating images are I2 and I3; the best quality means the largest field of view and uniform brightness distribution; (3b) Select 8 feature points on the three images R1, I2 and I3 that can represent the approximate direction distribution of the three images; (3c) Using the 2.5D Harris corner detection algorithm, non-maximum suppression windows are set respectively, where the radius of the three-dimensional non-maximum suppression window and the two-dimensional non-maximum suppression window are set to 25 and 20 respectively, and 700 feature points are extracted from the reference image R1; (3d) The generated 700 feature points are screened and 500 feature points are finally retained. 4. The method for fusion reconstruction of multi-view anisotropic three-dimensional images according to claim 1, characterized in that: the step (4) specifically comprises the following steps: (4a) Using the selected feature points, the coarse-precision three-dimensional affine transformation matrix from the floating image I2 to the reference image R1 and from the floating image I3 to the reference image R1 are calculated. The coarse-precision three-dimensional affine transformation matrix is used to perform a preliminary affine transformation of the two floating images I2 and I3 toward the reference image R1 to obtain the floating images I2’ and I3’. The basic directions of the three images R1, I2’ and I3’ are consistent. (4b) Input the reference image R1 and extract the feature points on the reference image R1 using the 2.5D Harris corner detection algorithm, as well as the floating images I2' and I3' after direction adjustment, and use the CLM coherent marker mapping matching algorithm, set the search radius to 10, and find the feature points on the floating images I2' and I3' that match the feature points on the reference image R1 through an iterative search; (4c) The inverse matrix of the coarse-precision three-dimensional affine transformation matrix is used to inversely transform the feature points on the floating images I2’ and I3’ onto the floating images I2 and I3, thereby obtaining the feature points on the floating images I2 and I3 that accurately match the reference image R1. Using these two sets of accurately matched feature point pairs, the affine transformation matrices for transforming the floating image I2 to the reference image R1 and the floating image I3 to the reference image R1 are calculated. 5. The method for fusion reconstruction of multi-view anisotropic three-dimensional images according to claim 1, characterized in that: the step (5) specifically comprises the following steps: (5a) Generate two full-space images with the same size as the reference image R1 and with a pixel value of 0, obtain the pixel value of each pixel point of the two full-space images, and calculate the position of each pixel point in the two full-space images in the two floating images I2 and I3 according to the affine transformation matrix; (5b) Linear interpolation is performed on the pixel values of the generated image using the pixel values of the eight neighboring positions around this position in the original image; if the position of the target image pixel in the original image exceeds the boundary, the pixel value of the part exceeding the boundary is taken as the pixel value of the reference image R1 at this position, and the generated image R2 is obtained based on the floating image I2, and the generated image R3 is obtained based on the floating image I3. 6. The method for fusion reconstruction of multi-view anisotropic three-dimensional images according to claim 1, characterized in that: the step (6) specifically comprises the following steps: (6a) The radius of the sliding one-dimensional window is selected as 5, and the three images R1, R2, and R3 are subjected to mean smoothing filtering at the three viewing angles of X, Y, and Z, respectively, to obtain three smoothed filtered images S1, S2, and S3; (6b) The three images R1, R2, and R3 are traversed pixel by pixel, and the brightness values of the generated images R2 and R3 are adjusted. After brightness matching, the pixel values at corresponding positions are R2(x, y, z) = R2(x, y, z)*(S1(x, y, z)/S2(x, y, z)), and R3(x, y, z) = R3(x, y, z)*(S1(x, y, z)/S3(x, y, z)), and the three images R1’, R2’, and R3’ with consistent brightness after brightness adjustment are obtained. 7. The method for fusion reconstruction of multi-view anisotropic three-dimensional images according to claim 1, characterized in that: the step (7) specifically comprises the following steps: (7a) Generate a completely empty image with the same size as the three images R1', R2' and R3' and with a pixel value of 0; (7b) For the full-space image of step (7a), traverse each pixel position, and for each pixel position (x, y, z) in the final output image _Io , take out the pixel values R'(x, y, z), R'(x+1, y, z), R'(x-1, y, z), R'(x, y+1, z), R'(x, y-1, z), R'(x, y, z+1), R'(x, y, z-1) of the corresponding positions and neighborhoods of the three images R1', R2' and R3'respectively; (7c) The three images are fused according to the three-dimensional gradient enhancement fusion method in the following manner: For the assignment of the pixel value of each pixel position in the final output image I _o , take

Where a=4b, b ranges from 0 to 1 and is adjusted according to the fusion effect. In the above formula, if I _o (x,y,z)<0 or I _o (x,y,z)>255, then

After traversing each time, the final output image I _o is obtained.

Images