CN-119937156-B - Focus depth reconstruction method based on Fourier laminated imaging

CN119937156BCN 119937156 BCN119937156 BCN 119937156BCN-119937156-B

Abstract

The invention discloses a method for reconstructing a large scene depth based on Fourier laminated imaging. The method utilizes the characteristic of the redundant information by utilizing the Fourier laminated imaging technology, establishes a layered reconstruction model, synthesizes the object information of each layer after accurately reconstructing the high-resolution object of each layer, and finally obtains the reconstruction result of large field of view, high resolution and large depth of field. Compared with the traditional Fourier laminated imaging technology, the reconstruction method of the invention remarkably expands the depth of field range of system imaging and is very suitable for pathological imaging of thick samples.

Inventors

ZHANG YUZHEN
Mo Yanlin
SUN JIASONG
CHEN QIAN
ZUO CHAO

Assignees

南京理工大学

Dates

Publication Date: 20260512
Application Date: 20250227

Claims (8)

1. A method for reconstructing a scene depth based on Fourier laminated imaging is characterized by comprising the following steps: the method comprises the steps of collecting original intensity images, namely using an annular LED plate as an illumination light source of a microscope, sequentially lighting three RGB (red, green and blue) lamp beads of each LED, taking light emitted by the lamp beads after the lamp beads irradiate a sample as plane monochromatic light, and collecting corresponding low-resolution bright field images of each lamp bead by synchronously triggering a camera to match with scanning of a lamp bead array; step two, determining a low-frequency cut-off frequency according to the numerical aperture of the microscope objective and the wavelength of the incident light, and calculating the spatial frequency of the incident light corresponding to each lamp bead according to the coordinate position of the LED lamp bead in the space; adding and averaging all shot low-resolution bright field images, then performing linear interpolation amplification to obtain an initial solution of a high-resolution image, and dividing the initial solution into a plurality of layers of complex-amplitude objects; setting defocus values corresponding to different layers of complex amplitude objects, for each layer of complex amplitude object, firstly converting each layer of complex amplitude referred to by the complex amplitude object into a frequency domain through Fourier transform, multiplying the frequency domain by a low-pass filter function corresponding to an oblique light angle and a transfer function corresponding to defocus amount to obtain each updated sub-aperture spectrum of each layer, and converting each updated sub-aperture spectrum of each layer into a space domain through inverse Fourier transform to obtain a complex amplitude object corresponding to each sub-aperture of each layer of different defocus positions; Step five, setting updating coefficients of each sub-aperture of each layer according to the distribution of the light intensity of each layer of complex amplitude object corresponding to the same sub-aperture; step six, iterative reconstruction, namely respectively transferring the complex amplitude object corresponding to each sub-aperture to a frequency domain by adopting a Fourier laminated imaging technology, and carrying out synthetic aperture calculation one by one in the frequency domain according to the update coefficient obtained by calculation in the step five, stopping iteration after the cost function of each layer is smaller than a set threshold value, and obtaining a high resolution result of each layer, otherwise, returning to the step four, wherein the specific method comprises the following steps: updating complex amplitude object corresponding to each sub-aperture of each layer by adopting Fourier laminated imaging technology At the same time, the update coefficient in the fifth step is multiplied by the index ; Wherein, the For the updated complex amplitude object corresponding to each sub-aperture of each layer, n represents the number of the lamp beads, To update the coefficients, P { } is bilinear interpolation upsampling, For a captured low resolution bright field image, For the complex amplitude object before update, m is the total number of captured images, The number of layers is the number of layers; the updated complex amplitude object of each sub-aperture of each layer Fourier transforming to the frequency domain and updating the complex amplitude object using the updated sub-aperture spectrum A portion of the lens corresponding to the sub-aperture; Wherein, the Representing an updated complex amplitude object, Representing the operation of a fourier transform, Representing an inverse fourier transform operation, The translation operation of the frequency spectrum is represented, the translation amount is determined by the angle of the oblique light corresponding to each lamp bead, As a function of the low-pass filtering, , For the spatial frequency of the signal to be transmitted, Is the number of waves to be used, Is a low-frequency cut-off frequency, The wavelength of the incident light; In order to update the step size, Defocus values corresponding to different layers of complex amplitude objects; Calculating a COST function COST, wherein the COST function COST is specifically as follows: the expression of the conjugate operation is given, The absolute value operation is represented by a function, P { } is bilinear interpolation upsampling as a cost function; Stopping iteration when the cost function of each layer is smaller than the set threshold value to obtain a high-resolution result of each layer, otherwise, keeping the complex amplitude object in the step six Replacing the complex amplitude object in the fourth step Returning to the fourth step, performing a new round of complex amplitude of each sub-aperture of each layer Is calculated and iterated; And step seven, synthesizing high-resolution results obtained by convergence of different layers to obtain a final large depth-of-field image.
2. The fourier stack imaging-based foreground depth reconstruction method of claim 1, wherein the specific formula for determining the low frequency cutoff frequency from the numerical aperture of the microscope objective and the wavelength of the incident light is: In the formula, Is a low-frequency cut-off frequency, The wavelength of the incident light is such that, Numerical aperture of the microscope objective.
3. The method for reconstructing the depth of view based on the fourier ptychographic imaging according to claim 1, wherein the specific method for calculating the spatial frequency of the incident light corresponding to each light bead according to the coordinate position of the LED light bead in the space is as follows: Calculating the spatial frequency of incident light corresponding to each lamp bead according to the coordinate position of the LED lamp bead in the space Wherein n represents the number of the lamp beads when At the time, the photographed image For bright field image, when At the time, the photographed image As dark field images, when At this time, the illumination is considered to be a matching illumination, the illumination aperture , Is the low frequency cut-off frequency.
4. The fourier stack imaging-based foreground depth reconstruction method of claim 1, wherein the high resolution image initial solution The method comprises the following steps: where P { } is bilinear interpolation upsampling, The low resolution bright field image is photographed, and m is the total number of photographed images.
5. The method for reconstructing a scene depth based on fourier ptychographic imaging according to claim 1, wherein in step three, an initial solution is performed Object divided into several layers of complex amplitude Each layer is an initial solution of the corresponding layer, and a calculation formula is as follows; In the formula, Is the number of layers.
6. The method for reconstructing a depth of view based on fourier ptychographic imaging of claim 5, wherein in step four, complex amplitude objects corresponding to each sub-aperture of different defocus positions are obtained The method comprises the following steps: wherein n represents the number of the lamp beads, For the number of layers of the material, Representing the operation of a fourier transform, Representing an inverse fourier transform operation, The translation operation of the frequency spectrum is represented, the translation amount is determined by the angle of the oblique light corresponding to each lamp bead, As a function of the low-pass filtering, , For the spatial frequency of the signal to be transmitted, Is the number of waves to be used, Is a low-frequency cut-off frequency, The wavelength of the incident light.
7. The method for reconstructing a scene depth based on fourier stack imaging as recited in claim 1, wherein the update coefficients are set according to a light intensity distribution of a complex amplitude object corresponding to each sub-aperture of each layer ; Wherein n represents the number of the lamp beads, For the number of layers of the material, In order to take the total number of images, For complex amplitude objects with different sub-apertures at different defocus positions, The expression of the conjugate operation is given, Representing an absolute value operation.
8. The method for reconstructing a large scene depth based on fourier ptychographic imaging according to claim 1, wherein in step seven, high resolution results obtained by convergence of different layers are combined to obtain a final large depth-of-field image The method comprises the following steps: Wherein, the For the updated complex amplitude results for each layer, Representing a conjugate operation.

Description

Focus depth reconstruction method based on Fourier laminated imaging Technical Field The invention belongs to optical microscopic imaging, and particularly relates to a method for reconstructing a large scene depth based on Fourier laminated imaging. Background In the traditional microscopic imaging field, there is always a pair of irreconcilable contradictions, namely between the imaging field of view and the imaging resolution. The image with larger field of view can be obtained by the low power mirror but the resolution is lower, and the image with high resolution can be obtained by the high power mirror but the field of view is smaller. The fourier stacked imaging technique is a computational microscopy imaging technique developed in recent years, and integrates the concepts of phase recovery and synthetic aperture, and performs alternate iteration according to light intensity information recorded in a space domain and a certain fixed mapping relationship in a frequency domain, so as to finally realize a large-view-field and high-resolution imaging result at the same time. In conventional fourier stack imaging systems, the sample is illuminated by plane waves at different angles and imaged by a low numerical aperture objective lens. The two-dimensional thin object is illuminated by plane waves from different angles, and the spectrum of the object is shifted to corresponding different positions on the back focal plane of the objective lens. Thus, some frequency components that would otherwise exceed the objective numerical aperture are translated into the objective numerical aperture and can be transferred to the imaging surface for imaging. In contrast, incident light with different angles can be equivalently overlapped pupil functions (sub-apertures) at different positions on a frequency spectrum, each time, a lamination is formed on a frequency domain through the frequency spectrums of the sub-apertures at different positions, a series of low-resolution images shot by a camera are then iterated in the frequency domain, spectrum information in the corresponding sub-apertures is sequentially updated, the sub-apertures and the sub-apertures overlap to expand the frequency domain bandwidth and recover high-frequency information exceeding the spatial resolution limit of an objective lens, and finally, a large-view high-resolution light intensity and a phase image of an object are simultaneously reconstructed. Thus, a low numerical aperture, low magnification objective lens is used to obtain a large field of view and high resolution imaging result. However, the conventional Fourier stack imaging model has a strong assumption that the object to be measured is a two-dimensional planar thin object. However, in practical situations, the sample to be observed is often a three-dimensional object with a thickness, and when such an object is reconstructed, errors such as artifacts are very likely to occur. Aiming at the problems, tian Lei and the like combine a classical Multi-Slice (Multi-Slice) model in the CT and nuclear magnetic resonance imaging fields with an FPM imaging method to construct a three-dimensional FPM reconstruction algorithm framework, and the super-resolution reconstruction （Tian L, Waller L. 3D intensity and phase imaging from light field measurements in an LED array microscope[J]. optica, 2015, 2(2): 104-111.）.Multi Slice model of a three-dimensional sample is realized by modeling the three-dimensional sample into a series of two-dimensional sample slices with specific intervals in a space domain, and all the sample slices are connected through free propagation of light waves in the forward modeling and reverse recovery processes. The algorithm can achieve high resolution reconstruction results for each layer for two resolution plates that overlap together. The left side super class proposes a Fourier laminated diffraction chromatography technology, expands a traditional FPM model into three dimensions, performs phase recovery and phase reconstruction （Zuo C, Sun J, Li J, et al. Wide-field high-resolution 3D microscopy with Fourier ptychographic diffraction tomography[J]. Optics and Lasers in Engineering, 2020, 128: 106003.）. in a three-dimensional Fourier domain, deduces a spectrum support domain of a microscopic system through the illumination numerical aperture and the objective numerical aperture of the microscopic system, obtains a specific structure of the support domain, and realizes an imaging method for carrying out Fourier lamination on a three-dimensional object. These methods often require extensive data acquisition and tomographic calculations to obtain high-precision three-dimensional reconstruction results. However, these techniques are difficult to apply to the pathology diagnosis industry because, on the one hand, doctors often do not need to obtain high-resolution three-dimensional information when making a diagnosis, but only need to obtain two-dimensional image information