A macroscopic Fourier stack super-resolution imaging method based on matrix scanning

By using matrix scanning and Fourier reconstruction algorithms in the macro imaging system, the problem of low resolution of the optical imaging system is solved, and super-resolution imaging in the macro imaging system is realized, reducing costs and expanding the application range.

CN114660060BActive Publication Date: 2024-11-29NANJING UNIV OF SCI & TECH
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202210224542.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-03-07
Publication Date
2024-11-29
Estimated Expiration
2042-03-07

AI Technical Summary

Technical Problem

The resolution of traditional optical imaging systems is limited by Abe's optical diffraction limit, making it difficult to improve image resolution under limited hardware conditions. Moreover, the application of Fourier stacked super-resolution imaging technology in macroscopic imaging systems is not yet mature and cannot be directly promoted. Long-distance imaging system.

Method used

Using the Fourier stacked super-segment imaging method based on matrix scanning, low-resolution images are collected through matrix scanning on the spectrum surface through imaging lens, and image processing is performed using SURF image matching algorithm and Fourier reconstruction algorithm to expand the imaging system. Effective caliber, realize super-resolution imaging of macroscopic imaging system.

Benefits of technology

Without the need for high-cost, large-diameter imaging lenses, the diffraction limit of the macroscopic imaging system is broken through, reducing the cost of achieving super-resolution imaging, solving the problem of difficulty in lighting angle transformation in traditional technology in macroscopic imaging systems, and expanding the Fu Application scope of Leaf laminated super-segment imaging technology.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114660060B_ABST
    Figure CN114660060B_ABST
Patent Text Reader

Abstract

The present invention discloses a macro Fourier stack super-resolution imaging method based on matrix scanning, comprising the steps of: 1) building a Fourier stack imaging system based on matrix scanning; 2) performing matrix scanning on the Fourier surface of the imaging system using a lens to acquire a series of low-resolution images under the original imaging lens aperture; 3) using a matrix scanning method based on SURF The present invention uses an imaging lens to perform matrix scanning in the far field to expand the spectrum range, and uses a Fourier stacking reconstruction algorithm to quickly reconstruct the amplitude information and phase information of the far-field target object, thereby obtaining a resolution that exceeds the diffraction limit of the system objective lens, effectively improving the imaging quality of the macro imaging system, and extending the traditional Fourier stacking imaging technology from micro to macro.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of computational imaging, and in particular relates to a macro Fourier stack super-resolution imaging method based on matrix scanning. Background Art

[0002] For optical imaging systems, the imaging process of the optical system can be regarded as a low-pass filter of information. Therefore, the resolution of traditional optical imaging systems is limited by the Abbe optical diffraction limit, and its lateral resolution is about 1 / 2 wavelength, which is only 200nm in the visible light band.

[0003] In order to improve the system resolution, the conventional approach is to increase the aperture of the imaging system to reduce the limitation of the Abbe diffraction limit, but the use of large-aperture lenses will increase the system development cost and extend the system development cycle, limiting its large-scale promotion and use. Therefore, how to sharpen the low-resolution image of the imaging system under limited hardware conditions to improve the image resolution is an urgent problem to be solved in the field of optical imaging.

[0004] In response to this demand, a series of novel microscopic imaging systems and methods have been proposed in the field of computational imaging in order to obtain super-resolution images with resolutions higher than the Abbe limit, which is the so-called "super-resolution" imaging technology, of which the most representative is Fourier Ptychographic Microscopy (FPM). This method illuminates the sample at multiple angles, records low-resolution images at different angles through a digital image receiver, and then uses a reconstruction algorithm to alternately iterate based on the mapping relationship between the spatial light intensity distribution information and the Fourier domain (i.e., frequency domain) angle information, ultimately obtaining the high-resolution intensity information and phase information of the sample, overcoming the restrictive relationship between the field of view and spatial resolution in traditional microscopic imaging systems, and achieving super-resolution imaging.

[0005] In the past 20 years, many research teams have achieved breakthroughs in the resolution limit of traditional microscopic imaging systems based on Fourier stacking super-resolution microscopy technology. In 2013, Zheng's team proposed a Fourier stacking microscopy system based on LED multi-angle illumination. The system uses an LED array to provide system illumination. Each LED lamp is in a different position, so the incident angle when irradiating the sample is also different. An image is taken for each incident angle, and then a high-resolution image can be restored from a series of low-resolution images using a reconstruction algorithm. In 2014, Williams et al. built a Fourier stacking microscopy (FPM) tumor cell counting system based on an LED array. In the same year, Chung et al. built an automatic white blood cell counting system also based on Fourier stacking microscopy. In 2015, Horstmeyer et al. successfully realized digital pathological analysis using a Fourier stacking microscopy system based on an LED array. In 2016, Pacheco et al. first realized a reflective Fourier stacking microscopy system that can be used for rapid large-field surface detection of samples.

[0006] The above-mentioned various Fourier stack-based super-resolution technologies are all studied from the perspective of microscopic imaging, and the application of this technology in macroscopic imaging systems is not yet mature. Due to the essential difference between the imaging distances of macroscopic and microscopic imaging systems, the traditional Fourier stack imaging technology is not applicable to long-distance imaging systems, and the method of spectrum scanning using illumination angles is not applicable to long-distance imaging systems. Therefore, when promoting this technology from micro to macro, it is not enough to simply scale up the model, but it is also necessary to modify and optimize the imaging system and reconstruction algorithm based on the imaging principle of the macroscopic imaging system. Summary of the invention

[0007] The purpose of the present invention is to provide a macro Fourier stack super-resolution imaging method based on matrix scanning, which can realize super-resolution imaging of long-distance imaging systems by simply combining matrix scanning acquisition with a reconstruction algorithm. It is suitable for improving the resolution of macro imaging systems and detecting targets.

[0008] The technical solution to realize the present invention is: a Fourier stack super-resolution imaging method based on matrix scanning, comprising the following steps:

[0009] Step 1: Set up a Fourier stack imaging system based on matrix scanning, and then go to step 2.

[0010] Step 2: Use the imaging lens to perform matrix scanning on the Fourier plane of the imaging system to acquire N*N low-resolution images under the original imaging lens aperture, whose imaging ranges overlap each other, N≥3, and then proceed to step 3.

[0011] Step 3: Select the low-resolution image at the center of the N*N low-resolution images as the reference image, and the remaining images as the images to be registered. Use the SURF-based image matching algorithm to perform image matching on the reference image and the images to be registered to obtain the matched N*N images, and then proceed to step 4.

[0012] Step 4: perform image cropping on the matched N*N images respectively to obtain cropped images, and proceed to step 5.

[0013] Step 5: Reconstruct the cropped image using the Fourier reconstruction algorithm to improve its resolution and obtain a restored image.

[0014] Compared with the prior art, the present invention has the following significant advantages: the present invention solves the problem of difficulty in changing the illumination angle when the traditional Fourier stack microscopy imaging technology is extended to a macroscopic imaging system, proposes to use a lens to scan the spectrum surface of the imaging system to achieve equivalent spectrum scanning, designs a macroscopic Fourier stack imaging system based on matrix scanning, obtains a series of conventional images under the original imaging lens aperture, and uses a reconstruction algorithm to iterate the obtained images, expands the equivalent aperture of the imaging system, and obtains high-definition images with improved image quality. It breaks through the diffraction limit of the macroscopic imaging system without the need for a high-cost large-aperture imaging lens, thereby reducing the cost of achieving super-resolution in the macroscopic imaging system. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flow chart of a macro Fourier stack super-resolution imaging method based on matrix scanning of the present invention.

[0016] Figure 2 It is a structural diagram of the macroscopic Fourier stack imaging system based on matrix scanning.

[0017] Figure 3 It is a schematic diagram of the light field propagation from the lens to the CCD detector.

[0018] Figure 4 It is a schematic diagram of the matrix scanning path.

[0019] Figure 5 It is a schematic diagram of the reconstruction algorithm principle. DETAILED DESCRIPTION

[0020] Based on the traditional Fourier stacking super-resolution microscopy imaging technology, the present invention extends the Fourier stacking super-resolution imaging technology from the micro to the macro, and organically combines the macro imaging system and the spectrum stacking scanning according to the basic principle of Fourier stacking, and proposes to use the lens to scan the spectrum surface of the imaging system to obtain a series of conventional images under the original imaging lens aperture, and use the reconstruction algorithm to iterate the obtained images to expand the equivalent aperture of the imaging system to obtain a restored image with improved resolution, and finally realize the super-resolution imaging of the macro imaging system, thereby expanding the application scope of the traditional Fourier stacking super-resolution imaging technology.

[0021] The following is combined with Figure 1 to Figure 5 The present invention further describes in detail a Fourier stack super-resolution imaging method based on matrix scanning.

[0022] Combination Figure 1 The macro Fourier stack super-resolution imaging method based on matrix scanning of the present invention comprises the following steps:

[0023] Step 1: Build a Fourier stack imaging system based on matrix scanning, such as Figure 2 As shown, it includes a laser, a beam expander, an aperture, a convex lens, a target, an imaging lens, and a CCD detector which are arranged in sequence on the same optical axis. The imaging lens is fixed on the CCD detector, and the CCD detector is fixed on the XY precision electric two-dimensional translation stage. The target is located in the far field of the imaging system. The laser uses a 532nm monochromatic laser. After the laser is expanded and filtered by a beam expander and an aperture, it provides relatively uniform illumination. A convex lens then converts the light field into a converging spherical wave to coherently illuminate the target, realizing far-field Fraunhofer diffraction at a limited distance. At this time, the light field emitted by the target, after long-distance propagation, the light field distribution on the front surface of the imaging lens is the far-field Fraunhofer diffraction result of the target.

[0024] Figure 3 The figure is a schematic diagram of the light field propagation from the lens to the CCD detector. After passing through the target object, the illumination light field is collected by the imaging lens and imaged on the target surface of the CCD detector. The imaging lens is equivalent to a Fourier transform, and the image captured by the CCD detector is the intensity information of the target object in the far field.

[0025] Go to step 2.

[0026] Step 2: Use the imaging lens to perform matrix scanning on the Fourier plane of the imaging system to acquire N*N low-resolution images under the original imaging lens aperture, and the imaging ranges thereof overlap with each other, N≥3, as follows:

[0027] Step 2-1, turn on the 532nm laser. The light emitted by the laser is expanded and filtered by a beam expander and an aperture to provide relatively uniform illumination. The light field is then converted into a converging spherical wave through a convex lens and shines on the target object. The power of the laser is adjusted until the illumination brightness is within the appropriate range to avoid overexposure or darkness.

[0028] Step 2-2: Power the CCD detector and set the exposure time, refresh rate, gain and other parameters of the CCD detector to ensure the quality of subsequent image acquisition.

[0029] Step 2-3: Use the host computer to send a signal to the XY precision electric two-dimensional translation stage, so that the XY precision electric two-dimensional translation stage drives the imaging device to move in the vertical plane. Figure 4 The matrix scanning shown has a moving step of 0.5 mm. The imaging lens first performs horizontal scanning from left to right in a plane perpendicular to the optical axis. The imaging range after each translation overlaps with the previous one in a certain proportion. After the first row of scanning is completed, the camera moves down one row and then scans from left to right until the scanning imaging of all positions is completed. N*N low-resolution images under the original imaging lens aperture are captured to obtain an original image sequence.

[0030] Step 2-4: Save the original sequence images on the host computer as the data set for the subsequent algorithm.

[0031] Go to step 3.

[0032] Step 3: Perform image matching on N*N low-resolution images (i.e., original sequence images), select the low-resolution image at the center as the reference image, and the remaining images as the images to be registered. Use the SURF-based image matching algorithm to perform image matching on the reference image and the images to be registered to obtain the matched N*N images, as follows:

[0033] Step 3-1: Perform the following processing on each low-resolution image:

[0034] Perform Gaussian smoothing at different scales:

[0035] L(x,y,σ)=G(x,y,σ)*I(x,y)

[0036]

[0037] Where σ is the Gaussian smoothing scale factor, L(x, y, σ) is the result of image smoothing, G(x, y, σ) is the Gaussian smoothing operator with a scale of σ, I(x, y) is the image intensity to be processed, and * is the convolution operation.

[0038] Through different Gaussian smoothing processes, a group of images with different smoothing processes are obtained, and this group of images is used as the bottom layer of the pyramid.

[0039] Then, each image in the bottom layer is downsampled to obtain the second and third layers, and then a complete image pyramid is obtained.

[0040] After obtaining the image pyramid, detect the point whose pixel value is larger or smaller than the surrounding 26 neighborhood points, and consider the point as a potential feature point for the next step of judgment and processing.

[0041] Since the edge of the image is more sensitive when detecting extreme points, the edge effect must be removed. After removing the edge points, a series of feature points are obtained, and the coordinate position (x, y) of the extreme points is obtained.

[0042] Step 3-2: Based on the coordinate position of the extreme point, the detected extreme point is described using the feature direction and generated feature descriptors.

[0043] The feature direction is determined based on the gradient size and gradient direction of each pixel. The gradients of the four pixels above, below, left and right of the pixel in the x and y directions are calculated, and the squares of each pixel are summed and then the square root is taken.

[0044] The gradient magnitude for each pixel is:

[0045]

[0046] Where m(x,y) is the gradient size of the current pixel, L(x,y) is the value of the current pixel, L(x+1,y) is the value of the pixel to the right of the pixel, L(x-1,y) is the value of the pixel to the left of the pixel, L(x,y+1) is the value of the pixel above the pixel, and L(x-1,y) is the value of the pixel below the pixel.

[0047] The gradient direction of each pixel is:

[0048]

[0049] Where θ(x,y) is the gradient direction of the current pixel, L(x,y) is the value of the current pixel, L(x+1,y) is the value of the pixel to the right of the pixel, L(x-1,y) is the value of the pixel to the left of the pixel, L(x,y+1) is the value of the pixel above the pixel, and L(x-1,y) is the value of the pixel below the pixel.

[0050] After calculating the gradient magnitude and the corresponding gradient direction, the gradient direction is used as the horizontal coordinate and the gradient magnitude as the vertical coordinate to generate a gradient histogram.

[0051] The peak with the largest ordinate in the histogram is taken as the main peak, and the corresponding gradient direction is the direction with the largest gradient, that is, the feature direction.

[0052] After the feature direction is determined, the coordinate axis is rotated to the feature direction to ensure the rotation invariance of the feature point. Then, with the feature point as the center, a 16×16 pixel area is taken and divided into 4×4 sub-areas. The gradient size and gradient direction in each area are calculated respectively, and the regional gradient histogram is generated to complete the description of the feature point.

[0053] Step 3-3: For the feature point A in the image to be registered, find a corresponding point B in the reference image as the reference point, and find the feature point C and the second closest point D in the reference image that are closest to point B by calculating the distance of the high-dimensional space vector.

[0054] Step 3-4: Calculate the distance between the reference point and the nearest feature point and the distance between the reference point and the next nearest feature point respectively, and calculate the ratio of the two. If the ratio is not less than the threshold, proceed to step 3-5; if the ratio is less than the threshold, proceed to step 3-6.

[0055] Step 3-5: Take point C in the reference image as the reference point, find the feature point C1 and the second closest point D1 in the reference image that are closest to the reference point by calculating the distance of the high-dimensional space vector, and return to step 3-4.

[0056] Step 3-6: The match is considered correct, and the matched N*N images are obtained respectively.

[0057] Go to step 4.

[0058] Step 4: Crop the matched N*N images into squares of 256*256 pixels respectively, remove the noise information at the edge of the image, increase the proportion of effective information in the image, reduce the calculation amount of the subsequent Fourier reconstruction algorithm, and obtain the cropped image.

[0059] Go to step 5.

[0060] Step 5: Reconstruct the cropped image using the Fourier reconstruction algorithm to improve its resolution and obtain a restored image. To ensure that the problem is solvable, the Fourier stacked reconstruction algorithm uses the intensity constraints and overlap rate constraints of the original image to reconstruct the object spectrum, iteratively updates the original image, obtains the synthetic spectrum, and obtains the restored image.

[0061] Figure 5 To reconstruct the algorithm flow chart, the details are as follows:

[0062] Step 5-1: Use the image sensor to obtain a series of low-resolution images I at different locations (m,n) , the image records the light intensity information of the sample.

[0063] By taking the root of the light intensity information, we can get the amplitude information at the corresponding scanning position.

[0064] The present invention selects the low-resolution image at the 0th row and 0th column position and interpolates it to obtain a high-resolution light intensity image, and then combines it with a high-resolution phase image with a phase of 0 to obtain an initial light field, which is used as the high-resolution initial solution to preliminarily constrain the result B(I (0,0) ).

[0065] Among them, B(…) represents bilinear interpolation, the subscript (0,0) represents the 0th row and 0th column, and I (0,0) Represents the low-resolution image at row 0, column 0.

[0066] Performing Fourier transform on it will give the corresponding spectrum information in the frequency domain, that is, the initial high-resolution spectrum:

[0067]

[0068] Among them, (u 0,0 ,v 0,0 ) is the frequency domain coordinate of the center of the spectral aperture corresponding to row 0 and column 0, u 0,0 represents the frequency domain horizontal axis, v 0,0 represents the ordinate in the frequency domain; O0 is the high-resolution spectrum of the initial solution, F{…} represents the Fourier transform of the complex amplitude, I (0,0) represents the low-resolution image at row 0, column 0, Represents the aperture function at row 0 and column 0.

[0069] Step 5-2: In the frequency domain, select a scanning position, use the spatial transfer function (OTF) of the imaging lens to constrain the initial high-resolution spectrum, intercept the spectrum information within the frequency domain sub-aperture of the position, and perform an inverse Fourier transform on the above spectrum information to obtain a low-resolution complex amplitude distribution in the spatial domain, that is, the target complex amplitude distribution:

[0070]

[0071]

[0072] Where, e represents the target spectrum and target complex amplitude to be updated; u represents the frequency domain horizontal coordinate, v represents the frequency domain vertical coordinate, (u, v) represents the frequency domain coordinate, represents the target spectrum to be updated at the position of the mth row and the nth column in the jth iteration, (u m,n ,v m,n ) represents the frequency domain coordinates of the aperture center in the spectrum corresponding to the position of the mth row and the nth column, represents the corresponding target complex amplitude distribution, P j (u,v) represents the aperture function of the mth row and nth column in the jth iteration, F -1{…} represents the inverse Fourier transform of the spectrum;

[0073] Step 5-3: Ensure that the phase remains unchanged, substitute the amplitude information of the target at the scanning position, and update the amplitude part of the target complex amplitude. The update formula is:

[0074]

[0075] Among them, c represents the actual information captured, represents the new target complex amplitude distribution of the scan position at the mth row and nth column in the jth iteration, Represents the low-resolution image at the mth row and nth column position, Represents the actual target complex amplitude corresponding to the position of the mth row and nth column in the jth iteration.

[0076] Step 5-4, perform Fourier transform on the updated target complex amplitude to obtain updated target spectrum information, substitute the target spectrum information into the spectrum components in the sub-aperture obtained previously, and complete the update of the spectrum at the scanning position. The update formula is:

[0077]

[0078]

[0079] in, represents the new target spectrum distribution of the mth row and nth column in the jth iteration, represents the new target complex amplitude distribution of the scan position at the mth row and nth column in the jth iteration, O j (uu m,n ,vv m,n ) represents the target spectrum information of the mth row and nth column in the jth iteration, Represents the actual spectrum information of the scan position at the mth row and nth column in the jth iteration.

[0080] Step 5-5, reselect the scanning position, and repeat steps 5-2 to 5-4 until the update of the complex amplitude information of all positions is completed, that is, the update of the spectrum information in all sub-apertures in the frequency domain is completed, and the first iteration is completed.

[0081] Step 5-6: Repeat steps 5-2 to 5-5 until the high-resolution complex amplitude of the object converges or reaches the preset number of iterations k. The spectrum at this time is the synthetic spectrum that can achieve super-resolution imaging. k represents the kth iteration.

[0082] Perform inverse Fourier transform on the synthesized spectrum and take the square of the modulus of the result to obtain the high-resolution image after image reconstruction:

[0083]

[0084] Among them I k represents the intensity of the restored image, represents the synthetic spectrum, k represents the kth iteration, F -1 {…} represents the inverse Fourier transform of the spectrum.

[0085] Based on the traditional Fourier stacking super-resolution microscopy imaging technology, the present invention extends the Fourier stacking super-resolution imaging technology from the micro to the macro, and organically combines the macro imaging system and the spectrum stacking scanning according to the basic principle of Fourier stacking, and proposes to use the lens to scan the spectrum surface of the imaging system to obtain a series of conventional images under the original imaging lens aperture, and use the reconstruction algorithm to iterate the obtained images to expand the equivalent aperture of the imaging system to obtain a restored image with improved resolution, and finally realize the super-resolution imaging of the macro imaging system, thereby expanding the application scope of the traditional Fourier stacking super-resolution imaging technology.

Claims

1. A Fourier stack super-resolution imaging method based on matrix scanning, characterized in that: The following steps are involved: Step 1: Build a Fourier stack imaging system based on matrix scanning: The matrix scanning-based Fourier stack imaging system comprises a laser, a beam expander, an aperture, a convex lens, a target, an imaging lens, and a CCD detector which are arranged in sequence on a common optical axis. The imaging lens is fixed on the CCD detector, and the CCD detector is fixed on an XY precision electric two-dimensional translation stage. The laser adopts a 532nm monochromatic laser. The light source is expanded and filtered by a beam expander and an aperture to provide relatively uniform illumination. A convex lens then converts the light field into a converging spherical wave to coherently illuminate the target, thereby realizing far-field Fraunhofer diffraction at a limited distance. At this time, the light field distribution of the light field emitted by the target on the front surface of the imaging lens after long-distance propagation is the far-field Fraunhofer diffraction result of the target. The target object of the matrix scanning-based Fourier stack imaging system is a transmission target object; Go to step 2; Step 2: Use an imaging lens to perform matrix scanning on the Fourier plane of the imaging system to acquire N*N low-resolution images under the original imaging lens aperture, and the imaging ranges thereof overlap with each other, and N≥3; The light field emitted by the target enters the imaging lens after long-distance propagation. Its long-distance propagation is regarded as a Fraunhofer far-field propagation. Its Fourier spectrum is intercepted by the aperture of the imaging lens and then received by the CCD detector to obtain a series of low-resolution images under the original imaging lens aperture, namely the original sequence image. The above images record the light intensity information of the target. The amplitude information at the corresponding scanning position is obtained by rooting the low-resolution image and sent to the host computer for subsequent calculation; Go to step 3; Step 3: Select the low-resolution image at the center of the N*N low-resolution images as the reference image, and the remaining images as the images to be registered. Use the SURF-based image matching algorithm to match the reference image and the images to be registered to obtain the matched N*N images, as follows: Step 3-1, select the low-resolution image at the center of the N*N low-resolution images as the reference image, and the remaining images as the comparison images; perform the following processing on each low-resolution image: perform Gaussian smoothing processing of different scales, establish an image pyramid, detect its extreme points, and obtain the coordinate positions of the extreme points; Step 3-2, based on the coordinate position of the extreme point, the detected extreme point is described using the feature direction and the generated feature descriptor; Step 3-3: For the feature point A in the image to be registered, find a corresponding point B in the reference image as the reference point, and find the feature point C and the second closest point D in the reference image that are closest to point B by calculating the distance of the high-dimensional space vector; Step 3-4, respectively calculate the distance between the reference point and the nearest feature point and the distance between the reference point and the next nearest feature point, and calculate the ratio of the two. If the ratio is not less than the threshold, proceed to step 3-5; if the ratio is less than the threshold, proceed to step 3-6; Step 3-5, taking point C in the reference image as the reference point, finding the closest feature point C1 and the second closest point D1 to the reference point in the reference image by calculating the distance of the high-dimensional space vector, and returning to step 3-4; Step 3-6: The match is considered correct, and the matched N*N images are obtained respectively; Go to step 4; Step 4: crop the matched N*N images respectively to obtain cropped images, and then proceed to step 5; Step 5: Reconstruct the cropped image using the Fourier reconstruction algorithm to improve its resolution and obtain a restored image.

2. The matrix scanning-based Fourier stack super-resolution imaging method according to claim 1, characterized in that: In step 4, the matched N*N images are cropped respectively to remove the edge noise information in the matched image, thereby increasing the proportion of effective information in the image and reducing the calculation amount of the subsequent Fourier reconstruction algorithm to obtain a cropped image.

3. The matrix scanning-based Fourier stack super-resolution imaging method according to claim 1, characterized in that: In step 5, the cropped image is reconstructed using a Fourier reconstruction algorithm to improve its resolution and obtain a restored image, as follows: Step 5-1: Interpolate the cropped image at the initial scanning position to obtain a high-resolution light intensity image, and then combine it with a high-resolution phase image with a phase of 0 to obtain the initial light field. Use the initial light field as the high-resolution initial solution to preliminarily constrain the reconstruction result. Perform Fourier transform on the initial light field to obtain the corresponding spectrum information in the frequency domain, that is, the initial high-resolution spectrum: Among them, (u 0,0 ,v 0,0 ) is the frequency domain coordinate of the center of the spectral aperture corresponding to row 0 and column 0, where u 0,0 represents the frequency domain horizontal axis, v 0,0 represents the ordinate in the frequency domain; O0 is the high-resolution spectrum of the initial solution, F{…} represents the Fourier transform of the complex amplitude, B(…) represents bilinear interpolation, I (0,0) represents the low-resolution image at row 0, column 0, represents the aperture function at row 0 and column 0; Step 5-2: In the frequency domain, select a scanning position, use the spatial transfer function (OTF) of the imaging lens to constrain the initial high-resolution spectrum, intercept the spectrum information within the frequency domain sub-aperture of the position, and perform an inverse Fourier transform on the above spectrum information to obtain a low-resolution complex amplitude distribution in the spatial domain, that is, the target complex amplitude distribution: Where, e represents the target spectrum and target complex amplitude to be updated; u represents the frequency domain horizontal coordinate, v represents the frequency domain vertical coordinate, (u, v) represents the frequency domain coordinate, represents the target spectrum to be updated at the position of the mth row and the nth column in the jth iteration, (u m,n ,v m,n ) represents the frequency domain coordinates of the aperture center in the spectrum corresponding to the position of the mth row and the nth column, represents the corresponding target complex amplitude distribution, P j (u,v) represents the aperture function of the mth row and nth column in the jth iteration, F -1 {…} represents the inverse Fourier transform of the spectrum; Step 5-3: Ensure that the phase remains unchanged, substitute the amplitude information of the target at the scanning position, and update the amplitude part of the target complex amplitude. The update formula is: Among them, c represents the actual information captured, represents the new target complex amplitude distribution of the scan position at the mth row and nth column in the jth iteration, Represents the low-resolution image at the mth row and nth column position, represents the actual target complex amplitude corresponding to the position of the mth row and the nth column in the jth iteration; Step 5-4, perform Fourier transform on the updated target complex amplitude to obtain updated target spectrum information, substitute the target spectrum information into the spectrum components in the sub-aperture obtained previously, and complete the update of the spectrum at the scanning position. The update formula is: in, represents the new target spectrum distribution of the mth row and nth column in the jth iteration, represents the new target complex amplitude distribution of the scan position at the mth row and nth column in the jth iteration, O j (uu m,n ,vv m,n ) represents the target spectrum information of the mth row and nth column in the jth iteration, Indicates the actual spectrum information of the scan position at the mth row and nth column in the jth iteration; Step 5-5, reselect the scanning position, and repeat steps 5-2 to 5-4 until the update of the complex amplitude information of all positions is completed, that is, the update of the spectrum information in all sub-apertures in the frequency domain is completed, and the first iteration is completed; Step 5-6, repeat steps 5-2 to 5-5 until the high-resolution complex amplitude of the target object converges. The complex amplitude at this time is the optimal solution of the high-resolution complex amplitude of the target object, and the image reconstruction is completed to obtain a restored image.

Citation Information

Patent Citations

  • Method and device for eliminating black-matrix effect of phase type spatial light modulator

    CN101794029A

  • Continuous terahertz wave Fourier laminated microscopic imaging system and method

    CN111307759A