CN106204434A - A kind of Image Iterative reconstructing method towards large visual field high resolution micro-imaging - Google Patents

Abstract

The invention discloses an image iterative reconstruction method for large-field-of-view and high-resolution microscopic imaging. Firstly, an LED array is used as the illumination source of the microscope, and each LED element is sequentially lit, and the corresponding image is collected after irradiating the sample. ; Initialize the amplitude and phase of the high-resolution image from the low-resolution image taken by illuminating the sample with the centrally located LED element in the LED array; use the incremental gradient method to step through each acquired image one by one in the frequency domain Synthetic aperture operation; update the incremental gradient iteration coefficient with the cost function value as the criterion; stop iteration when the incremental gradient iteration coefficient is less than a given threshold. The invention does not require complex parameter adjustment, has strong resistance to noise in the collected images, and can very stably and accurately reconstruct high-resolution images with a large field of view.

Description

An Image Iterative Reconstruction Method for Large Field of View and High Resolution Microscopic Imaging

technical field

The invention belongs to optical microscopic imaging technology, in particular to an image iterative reconstruction method for large field of view and high resolution microscopic imaging.

Background technique

In the field of microscopic imaging, higher resolution has always been the goal of pursuit, but there is a key problem while improving the resolution, that is, the spatial bandwidth product of the microscope does not increase with the resolution. From the perspective of the imaging system, in order to achieve high resolution, the numerical aperture of the microscope objective lens must be increased, but the improvement of spatial resolution and the expansion of the field of view are often a pair of contradictions that are difficult to reconcile. In short, the whole picture of the inspected object can be seen under the low magnification lens, but only a small part of the inspected object can be seen when it is replaced with a high magnification objective lens. At present, in order to break through the contradiction between the resolution and the size of the field of view, the common method is to use a conventional microscope system with high-precision mechanical scanning and a later spatial image stitching method to stitch and fuse multiple high-resolution images of small fields of view to generate a large image. High-resolution images of the field of view ([1] 2013205777012, a device suitable for stitching images of Mycobacterium tuberculosis acid-fast staining). However, due to the introduction of mechanical moving devices, the imaging stability and imaging speed of the system have become a pair of contradictions that are difficult to reconcile. Improving the scanning speed will definitely affect the imaging stability. Therefore, in order to break through the contradiction between resolution and field of view and not introduce mechanical moving devices, computational imaging methods proposed in recent years must be adopted, such as imaging methods based on synthetic aperture.

The scanning imaging method based on the principle of synthetic aperture imaging was first proposed by Hoppe to study the crystal structure, and the effectiveness of this method was verified by studying the scanning transmission electron diffraction microscopy imaging of crystals and amorphous materials. Rodenburg and Faulkner combined the phase recovery algorithm to improve this method for many times. At present, this imaging method has been experimentally confirmed in different wavebands such as visible light domain, X-ray, and electron microscope, and several technologies have been developed to improve imaging quality and resolution. rate, the technology shows great potential in large-format imaging and high-resolution imaging. The traditional synthetic aperture imaging technology is to irradiate the incident plane wave to different parts of the sample by moving a fully transparent small hole (or the sample itself), that is, the size, geometry and position of the illumination beam are controlled by the small hole, and use A series of diffraction intensity patterns thus obtained can reconstruct the amplitude and phase information of the sample to be measured ([2] Wang Yali et al. Research on the key parameters of the illumination beam in visible light domain lamination imaging[J]. Acta Physica Sinica, 2013, Vol .62, No. 6.064206-1-9). The key to synthetic aperture imaging is that each time a "sub-aperture" of the sample to be tested is irradiated, that is, a certain part of the sample to be tested, it must overlap with at least one other "sub-aperture". In this way, a reconstruction algorithm can be established to satisfy the constraints of the diffraction distribution of other “sub-apertures” when reconstructing the complex amplitude of each “sub-aperture” respectively, so that the overall complex amplitude information of the final sample to be measured is all The common solution of "sub-apertures", so that each "sub-aperture" is spliced into a large field of view and high-resolution image of the sample to be tested. Synthetic aperture imaging can be said to be a robust and simple microscopic imaging technique, but it lacks a robust image reconstruction algorithm. Especially when the signal-to-noise ratio of the collected image is low, it is often difficult to obtain an ideal reconstructed image. Therefore, how to improve the reconstruction quality and signal-to-noise ratio has become a technical problem that must be overcome in synthetic aperture imaging technology.

Contents of the invention

The purpose of the present invention is to provide an image iterative reconstruction method for large field of view and high resolution microscopic imaging, so as to solve the contradiction between the resolution of the microscopic system and the size of the field of view, and improve the spatial bandwidth product of the microscopic system .

The technical solution to realize the purpose of the present invention is: an image iterative reconstruction method for large-field-of-view high-resolution microscopic imaging, the steps are as follows:

Step 1, image acquisition: the LED array is used as the illumination source of the microscope, and each LED element is sequentially lit, and the corresponding image is collected after irradiating the sample;

Step 2, initialization: initialize the amplitude and phase of the high-resolution image by using the low-resolution image captured by the LED element in the center of the LED array to irradiate the sample;

Step 3, iterative reconstruction: use the incremental gradient method to perform synthetic aperture operations on each image collected in the frequency domain;

Step 4, update the incremental gradient iteration coefficient: update the incremental gradient iteration coefficient with the cost function value as the criterion;

Step 5, stop the iteration judgment: when the incremental gradient iteration coefficient is less than a given threshold, stop the iteration, and the amplitude and phase of the high-resolution image at this time are the finally obtained large-field high-resolution microscopic image.

In the present invention, the illumination sources in step 1 are red light, green light or blue monochromatic light to illuminate the object, and for each illumination wavelength, the above five steps are used to reconstruct the image, and the reconstructed three The group images are respectively synthesized as the red, green and blue components of the final true color image to obtain a reconstructed true color image.

Compared with the prior art, the present invention has significant advantages: (1) The LED array is used to generate incident light at different angles to irradiate the sample to be tested, and then multiple low-resolution images are taken with a camera, and finally the frequency-domain synthetic aperture technology is used to combine multiple A large field of view low resolution image is synthesized into a large field of view high resolution image without any mechanical scanning device, and the system has high imaging stability. Compared with the high-precision mechanical scanning and post-space image stitching methods of conventional microscope systems, the cost of the system is greatly reduced. (2) It does not require complex parameter adjustment, and has a strong resistance to noise in the collected images, and can reconstruct large-field-of-view high-resolution images very stably and accurately.

The present invention will be described in further detail below in conjunction with the accompanying drawings.

Description of drawings

Figure 1 is a schematic diagram of the optical path of a microscope based on a programmable LED array.

Fig. 2 is a schematic diagram of the coordinate system of each pixel in the LED array.

FIG. 3 is a schematic flowchart of the image iterative reconstruction method for large-field-of-view and high-resolution microscopic imaging of the present invention.

Figure 4(a) is a low-resolution image captured by using the LED element in the center of the LED array to illuminate the 1951USAF resolution plate sample under a 4x objective lens (numerical aperture 0.1); select the image in the box of Figure 4(a) Directly zoom in and get Figure 4(b). Then choose a smaller area to zoom in directly, and get Figure 4(c).

Figure 5(a) is a large-field high-resolution microscopic image reconstructed on the 1951USAF resolution plate using the method of the present invention; the image in the box in Figure 5(a) is selected to be directly enlarged, and Figure 5(b) is obtained. Then choose a smaller area to zoom in, and get Figure 5(c).

detailed description

The hardware platform on which the image iterative reconstruction method for large-field-of-view and high-resolution microscopic imaging depends is a microscope based on a programmable LED array. Combined with FIG. 1 , the microscope based on the programmable LED array mainly includes a microscope objective lens 1 , a sample 2 , a stage 3 , and an LED array 4 . The LED array 4 is used as the illumination source of the microscope, which is directly placed under the sample stage 3, and the distance H from the upper surface of the stage is generally between 20-100 mm, and the center of the LED array 4 is located at the microscope objective lens 1 on the optical axis. The LED array 1 includes several LED elements 5 arranged regularly to form a two-dimensional array. Among them, the single LED elements are all red, green and blue three-color LEDs, and their typical wavelengths are red light 635nm, green light 525nm and blue light 475nm. The center distance d between each LED element is typically 1-10mm. The LED array 4 does not need to be processed separately, and can be directly purchased in the market. It includes a group of multiple LEDs arranged in an array, and these LEDs are physically and electrically connected through a fixed substrate. Table 1 shows the product parameters of a commercially available LED array. In this LED array, there are 32 rows and 32 columns of LED elements, a total of 1024, and the brightness of a single LED is above 2000cd/m ² .

Table 1 Physical parameters of LED array

Each LED element in the LED array 1 can be individually lit by realizing that the specific method of lighting the LED element is the prior art, and the implementation circuit can adopt (but not limited to) existing technologies such as single-chip microcomputer, ARM, or programmable logic devices. It can be realized, and the specific implementation method can refer to relevant literature, such as Guo Baozeng, Deng Chunmiao: FPGA-based LED display control system design [J]. Liquid Crystal and Display, 2010,25(3):424-428.

Before the implementation of the image iterative reconstruction method for large-field-of-view and high-resolution microscopic imaging in the present invention, the spatial frequency of the illumination light corresponding to each LED element in the LED array 4 is first marked, and the specific method is as follows: Combined with FIG. 2, establish Coordinate System. The rectangular area represents the effective area of the LED array 4, and the coordinate origin is located at the center of the optical axis of the imaging system. For any LED element P, whose position coordinates are (P _x , P _y ), first calculate the spatial frequency vector of the illumination light corresponding to the LED element:

${\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; \&pi;</span>}}{\mathrm{<span\; class="notranslate">\; \&lambda;</span>}}<span\; class="notranslate">\; (</span>\frac{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}}{\sqrt{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; h</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}}<span\; class="notranslate">\; ,</span>\frac{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}{\sqrt{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; h</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}}<span\; class="notranslate">\; )</span>$

Similarly, for each LED element in the LED array 4, its corresponding spatial frequency vector can be calculated, denoted as u _i , subscript i=1, 2,..., N where N is the element in the LED array 4 The total number of LED elements, λ is the wavelength of monochromatic illumination light (red, green or blue). u _i is the spatial frequency vector of the illumination light of the i-th LED.

In conjunction with Fig. 3, based on the above-mentioned technology, the present invention is oriented to an image iterative reconstruction method for large-field-of-view high-resolution microscopic imaging, the steps of which include:

Step 1: Image acquisition. The LED array 4 is used as the illumination source of the microscope, sequentially lights up each LED pixel, and collects corresponding images after irradiating the sample. Since the entire LED array contains N LED pixels, a total of N low-resolution images are taken, denoted as I _i (r), i=1,2,...,N, r is the two-dimensional coordinate r of the real space =(x,y). Ideally, these captured low-resolution images I _i (r) can be expressed as

in is the two-dimensional Fourier transform, u is the spatial frequency coordinate corresponding to r, is the two-dimensional inverse Fourier transform, O(u) is the Fourier transform of the high-resolution object function to be found, namely P(u) is the aperture function of the microscope. It can be seen that the Fourier transform of the object function is firstly moved by the spatial frequency vector u _i of the illumination light of the i-th LED, and then limited by the aperture function P(u) of the microscope, after the inverse Fourier transform in the real A low-resolution image I _i (r) is formed spatially, which corresponds to N images I _i (r) captured by the camera, i=1, 2, . . . , N.

Solving the Fourier transform O(u) of the high-resolution object function is essentially equivalent to solving the following optimization problem

ε is a cost function that measures the difference between the captured image and the computed image. The above formula shows that the Fourier transform O(u) of the high-resolution object function to be found is the solution to minimize the cost function ε. In the following, for the convenience of description, all the above formulas and variables are expressed in vector form: Represents a column vector in which the captured low-resolution image I _i (r) is arranged in pixel order, and the number of elements in the column vector is equal to the number of pixels in the image, both of which are M. I _i is a low-resolution image vector. The Fourier transform O(u) representing the high-resolution object function is a column vector arranged in pixel order, the number of elements in the column vector is L, and O is the Fourier transform vector of the high-resolution object function. Since the resolution of the reconstructed image is often much higher than that of the original image, L>>M. P _i is an M×L matrix determined by the aperture function P(u), called the aperture function matrix, and its function is to extract from the M elements of the Fourier transform vector of the entire high-resolution object function L elements bounded by the aperture. After expressed by the above vector, the cost function ε can be expressed as

$\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>\sqrt{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; f</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; o</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>{<span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>\sqrt{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; g</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>{<span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}$

where ||·|| represents the Euclidean norm, F is the matrix expression of the discrete Fourier transform, and g _i ≡ F ^-1 P _i O. According to Parseval's theorem, the cost function ε can be simplified as

$\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; \&Psi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; o</span>}<span\; class="notranslate">\; |</span>{<span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}$

in

${\mathrm{<span\; class="notranslate">\; \&Psi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; \&Pi;</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&Right\; Arrow;</span>\sqrt{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; o</span>}<span\; class="notranslate">\; )</span>$

represents the updated sub-spectrum; Represents the real-space modular projection operation

${\mathrm{<span\; class="notranslate">\; \&Pi;</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&Right\; Arrow;</span>\sqrt{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; o</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; f</span>}\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; a</span>}\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; g</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; g</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}{<span\; class="notranslate">\; |</span>}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span>\sqrt{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}$

It represents that the magnitude of gi is partly taken from the measured value replaced, while keeping its phase part unchanged. Diag() represents a diagonal matrix.

Step 2: Initialization. The amplitude and phase of the high-resolution image are initialized by using the low-resolution image captured by the LED element in the center of the LED array to illuminate the sample, as shown in the following formula:

${\mathrm{<span\; class="notranslate">\; o</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; f</span>}\sqrt{<span\; class="notranslate">\; \&Up\; Arrow;</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; r</span>}}}$

In the formula, I _center is the low-resolution image vector captured by the LED element in the center of the LED array 4 irradiating the sample; ↑ represents image upsampling, that is, the image is interpolated from the original M pixels to L pixels. The interpolation method here can directly adopt the existing technology, such as the nearest neighbor interpolation method, bilinear interpolation method or bicubic interpolation method, etc., and the selection of the interpolation method will not affect the final reconstruction result. The incremental gradient iteration coefficient α ⁰ is initialized to α ⁰ =1, and the cost function value ε ⁰ is initialized to Here the subscript 0 of O ₀ represents the number of iterations of the inner loop, and the superscript 0 of α ⁰ and ε ⁰ represents the number of iterations of the outer loop.

Step 3: Iterative reconstruction. The incremental gradient method is used to perform synthetic aperture operation on each acquired image one by one in the frequency domain. In order to minimize the cost function ε, the cost function is written in incremental form, namely:

ε _i ＝||Ψ _i -P _i O|| ²

Therefore, the incremental gradient of the cost function can be expressed as

$<span\; class="notranslate">\; \&dtri;</span>{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}^{<span\; class="notranslate">\; *</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&Psi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; o</span>}<span\; class="notranslate">\; )</span>$

Then, the incremental gradient method is used to update the Fourier transform vector of the high-resolution object function, which can be subdivided into three sub-steps:

In the first step, the Fourier transform vector of the high-resolution object function is updated using the incremental gradient method as follows:

${\mathrm{<span\; class="notranslate">\; o</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; o</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&dtri;</span>{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; o</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}^{<span\; class="notranslate">\; *</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&Psi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; o</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; )</span>$

Among them, α ^k is the incremental gradient iteration coefficient; Represents the sub-spectrum updated by I _i

${\mathrm{<span\; class="notranslate">\; \&Psi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; \&Pi;</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&Right\; Arrow;</span>\sqrt{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; o</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; )</span>$

The subscript i in the above formula represents the number of iterations of the inner loop, and the superscript k represents the number of iterations of the outer loop. W _i is the weight matrix

${\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; a</span>}\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}{<span\; class="notranslate">\; |</span>}_{\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; a</span>}\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>$

Diag() represents a diagonal matrix, and its diagonal elements are in order |P _m,i | _max represents the maximum value of the aperture function modulus. The function of W _i is to weight the frequency components inside the aperture function and exclude the frequency components outside the aperture function.

In the second step, the Fourier transform vector of the high-resolution object function After the update is completed, the number of iterations of the inner loop is increased by 1, that is, i←i+1, switch to the next low-resolution image vector I _i , i=1,2,...,N, and repeat the first step.

In the third step, after all the images have been traversed once, that is, each low-resolution image vector I _i , i=1, 2, .

Step 4: Incremental gradient iteration coefficient update, divided into three sub-steps:

1. Calculate the Fourier transform vector of the current high-resolution object function The error function corresponding to the corresponding cost function value ε ^k :

${\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>\sqrt{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; f</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; o</span>}}_{\mathrm{<span\; class="notranslate">\; N</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>{<span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}$

2. Use the cost function value ε ^k as the criterion to update the incremental gradient iteration coefficient. The update formula is as follows:

The above formula shows that the incremental gradient iteration coefficient α ^k is reduced to half of the original value when the cost function value ε ^k no longer decreases. This can ensure the adaptive convergence of the algorithm and effectively improve the robustness of the algorithm to noise.

3. The number of iterations of the outer loop is increased by 1, that is, k←k+1, and the number of iterations of the inner loop is cleared to 0, that is, i←0.

The core here is that the incremental gradient iteration coefficient α ^k tends to decrease gradually to ensure the stable convergence of the algorithm.

Step 5: Stop iterative judgment. When the incremental gradient iteration coefficient α ^k is less than a given threshold (the recommended value is 0.001), the iteration is stopped. At this time, the Fourier transform vector of the high-resolution object function obtained by iteration Perform an inverse Fourier transform into the spatial domain

${\mathrm{<span\; class="notranslate">\; o</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; u</span>}\mathrm{<span\; class="notranslate">\; l</span>}\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; f</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; o</span>}}_{\mathrm{<span\; class="notranslate">\; N</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}$

That is, the amplitude distribution |O _result | and the phase distribution angle(O _result ) of the large field of view and high resolution of the sample to be tested are obtained, where angle() represents the argument of a complex number. If the stop condition is not met, go back to step 3 and perform an inner loop again until the stop condition is met.

The method for automatically determining the focal plane of an object in the present invention is as follows: before step 1 of an image iterative reconstruction method for large-field-of-view and high-resolution microscopic imaging: before image acquisition. It is necessary to ensure that the height of the stage 3 is correct, that is, the sample 2 can be accurately located on the focal plane of the microscopic objective lens 1, and for the microscopic objective lens 1, it is an infinity objective lens, that is, to ensure that the object is strictly located on the front focal plane of the microscopic objective lens 1 , at this moment, the microscope objective lens 1 can image the sample 2 clearly. In order to ensure that the height of the stage 3 is correct, follow the steps below to adjust the height of the stage:

①The LED array 4 is used as the illumination source of the microscope to light up all the LED pixels therein.

② Adjust the height of the stage 3 to the minimum value z ₀ , and collect the corresponding images

③ Solve according to the following formula Corresponding defocus measurement function value

${\mathrm{<span\; class="notranslate">\; Laplace</span>}}_{{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; I</span>}}_{{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; +</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; I</span>}}_{{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span>$

${\mathrm{<span\; class="notranslate">\; L</span>}}_{{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; \&Sigma;\&Sigma;Laplace</span>}}_{{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; for\&Sigma;\&Sigma;Laplace</span>}}_{{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; T</span>}$

Where step is the gradient solution step size, generally 1-5 pixels. T is the gradient threshold, generally taken as 3% of the maximum gray value of the image.

④ Raise the height of stage 3 to z ₁ and collect corresponding images

⑤Similarly, solve The corresponding defocus measurement function

⑥Repeat the above process until the minimum defocus measurement function value in all ranges is found At this time, the corresponding height of the stage 3 is the correct value. At this time, it can be ensured that the microscope objective lens 1 can clearly image the sample 2 .

Image sequence selection for iterative reconstruction in Step 3 of the present invention: There are two alternatives here, one is that the LEDs are updated sequentially from the center (on the optical axis) to the periphery (away from the optical axis). Another is to update in descending order according to the intensity value ||I _i || of the image. Both update orders can get good refactoring results.

The above steps are only for monochromatic lighting (such as red, green, blue, etc.), if you want to reconstruct a true color image. Then each LED element needs to use red light, green light, and blue light to illuminate the object, and then use the above five steps for image reconstruction for each illumination wavelength, and the three reconstructed images are respectively used as the final true color image The red, green and blue components can be synthesized.

In order to test the image iterative reconstruction method of large-field-of-view high-resolution microscopic imaging, we selected the 1951USAF resolution test board for imaging testing. In the experiment, the LED array used contains a total of 441 LEDs in 21 rows and 21 columns, and it is used to generate 225 different angles of illumination light. The spacing between the LEDs is 2.5mm, and the emitted red light has a wavelength of 632.8nm. The numerical aperture of the microscope objective used in the system is 0.1, and the magnification is 4x. A low-resolution image taken of the sample illuminated by the centrally located LED element of the LED array is shown in Figure 4. Select the image in the box in Figure 4(a) and directly zoom in to get Figure 4(b). Then choose a smaller area to zoom in directly, and get Figure 4(c). As can be seen from the figure, the smallest resolvable feature is the third element of the seventh group. According to the physical parameters of the 1951USAF resolution test board (see Table 2), the original imaging resolution of the imaging system is about 6.2 μm/line pair. This is in good agreement with the results inferred from the Rayleigh diffraction limit formula for imaging systems. However, the high-resolution image reconstructed by the present invention is shown in FIG. 5 , and the image in the box in FIG. 5( a ) is selected and directly enlarged to obtain FIG. 5( b ). Then choose a smaller area to zoom in directly, and get Figure 5(c). The smallest feature in the resolution board can be resolved (third element of the ninth group). It can be seen from Table 2 that the resolution of the combined imaging system is better than 1.56 μm/line pair, that is, the resolution is higher than 1000 um/(645*2)=0.775 um. Comparing Fig. 5(c) and Fig. 4(c), it can be clearly seen that the method of the present invention can realize large-field-of-view high-resolution imaging in the case of low numerical aperture and large field of view, and the reconstructed image has a good signal-to-noise ratio.

Table 2 Physical parameters of 1951USAF resolution test board

Claims (10)

1. the Image Iterative reconstructing method towards large visual field high resolution micro-imaging, it is characterised in that step is as follows:

Step one, image acquisition: LED array, as microscopical lighting source, sequentially lights each of which LED element, shine Corresponding image is gathered after penetrating sample；

Step 2, initializes: utilize the LED element being positioned at center in LED array to irradiate the low resolution figure taken by sample As initializing amplitude and the phase place of high-definition picture；

Step 3, iterative reconstruction: use incremental gradient method to carry out the every piece image gathered the most one by one synthesizing hole Footpath computing；

Step 4, incremental gradient iteration coefficient updates: be updated incremental gradient iteration coefficient with cost function value for criterion；

Step 5, stops iteration and judges: when incremental gradient iteration coefficient is less than a given threshold value, stops iteration, now Amplitude and the phase place of high-definition picture be exactly the large visual field high resolution micro-image finally given.

2. the Image Iterative reconstructing method towards large visual field high resolution micro-imaging, it is characterised in that step is as follows:

Step one, image acquisition: LED array, as microscopical lighting source, sequentially lights each of which LED element, shine Corresponding image is gathered after penetrating sample；

Step 2, initializes: utilize the LED element being positioned at center in LED array to irradiate the low resolution figure taken by sample As initializing amplitude and the phase place of high-definition picture；

Step 3, iterative reconstruction: use incremental gradient method to carry out the every piece image gathered the most one by one synthesizing hole Footpath computing；

Step 4, incremental gradient iteration coefficient updates: be updated incremental gradient iteration coefficient with cost function value for criterion；

Step 5, stops iteration and judges: when incremental gradient iteration coefficient is less than a given threshold value, stops iteration, now Amplitude and the phase place of high-definition picture be exactly the large visual field high resolution micro-image finally given；

Step 6, the respectively HONGGUANG of the lighting source in step one, green glow, the monochromatic light of blue light illuminate object, for often Kind illumination wavelengths is respectively adopted above-mentioned five steps and carries out image reconstruction, using three groups of images of each via Self-reconfiguration as final The red, green, blue component synthesis of true color image obtains reconstructing true color image.

Image Iterative reconstructing method towards large visual field high resolution micro-imaging the most according to claim 1 and 2, it is special Levy before being that iterative reconstruction is implemented, first the spatial frequency of the illumination light corresponding to LED element every in LED array is marked Note, concrete grammar is as follows: set up coordinate system, and wherein rectangular area represents the effective coverage of LED array (4), and zero is positioned at The optical axis central authorities of imaging system；For any one LED element P, its position coordinates is (P_x,P_y), first calculate this LED element The spatial frequency vector of corresponding illumination light:

$u_i=(u_x,u_y)=\frac{2\mathrm{\&pi;}}{\mathrm{\&lambda;}}(\frac{P_x}{\sqrt{P_x^2+P_y^2+H^2}},\frac{P_y}{\sqrt{P_x^2+P_y^2+H^2}})$

It is calculated its corresponding spatial frequency vector for LED element each in LED array 4, is denoted as u_i, subscript i=1, 2 ..., N, total number of LED element during wherein N is LED array 4, λ is the wavelength of illumination light, u_iIllumination light for i-th LED Spatial frequency vector.

Image Iterative reconstructing method towards large visual field high resolution micro-imaging the most according to claim 1 and 2, it is special Levying and be in step one, the enforcement step of image acquisition is: LED array (4), as microscopical lighting source, sequentially lights it In each LED pixel, irradiate after sample and gather corresponding image；Owing to whole LED array comprising N number of LED pixel altogether, Amount to shooting N width low-resolution image, be denoted as I_i(r), i=1,2 ..., N, r be the real space two-dimensional coordinate r=(x, y), will The low-resolution image I photographed_iR () column vector that according to pixels order is arranged into, obtains low-resolution image vector I_i。

Image Iterative reconstructing method towards large visual field high resolution micro-imaging the most according to claim 1 and 2, it is special Levy and be in step 2, initialized enforcement step by: utilize LED array is positioned at center LED element irradiate sample clapped The low-resolution image taken the photograph is to initialize amplitude and the phase place of high-definition picture, i.e.

$O_0=F\sqrt{\&UpArrow;I_{center}}$

In formula, I_centerFor LED array 4 is positioned at the LED element at center irradiate low-resolution image taken by sample to Amount；↑ representative image up-samples, and will image is M pixel by original L picture element interpolation；Incremental gradient iteration coefficient α⁰Initially Turn to α⁰=1, cost function value ε⁰It is initialized asO₀Subscript 0 represent interior loop iteration number of times, α⁰With ε⁰Upper Mark 0 represents outer circulation iterations.

Image Iterative reconstructing method towards large visual field high resolution micro-imaging the most according to claim 1 and 2, it is special Levying and be in step 3, iterative reconstruction is divided into three sub-steps:

The first step, uses incremental gradient method to update the Fourier Transform vector of high-resolution thing function as follows:

$O_{i+1}^k=O_i^k+{\mathrm{\&alpha;}}^k{W}_i\&dtri;{\mathrm{\&epsiv;}}_i^k=O_i^k-{\mathrm{\&alpha;}}^k{W}_i{P_i}^{*}({\mathrm{\&Psi;}}_i^k-P_i{O}_i^k)$

Wherein, α^kFor incremental gradient iteration coefficient；Represent by I_iSub-frequency spectrum after renewal:

${\mathrm{\&Psi;}}_i^k={\mathrm{\&Pi;}}_{m\&RightArrow;\sqrt{I_i}}\left(P_i{O}_i^k\right)$

Subscript i in above formula represents interior loop iteration number of times, and subscript k represents outer circulation iterations；W_iFor weight square Battle array:

$W_i=Diag\left({\left|P_{m,i}\right|}_{max}^{-2}\right)$

Diag () represents diagonal matrix, and its diagonal entry is followed successively by|P_{M, i}|_maxRepresent the maximum of aperture function mould Value；

Second step, the Fourier Transform vector of high-resolution thing functionUpdate complete after, interior loop iteration number of times adds 1, i.e. I ← i+1, is switched to next width low-resolution image vector I_i, i=1,2 ..., N, repeats the first step；

3rd step, after all image traversal are crossed once, the most every width low-resolution image vector I_i, i=1,2 ..., N both participates in Interative computation is until i=N, and step 3 completes.

Image Iterative reconstructing method towards large visual field high resolution micro-imaging the most according to claim 1 and 2, it is special Levying and be in step 4, incremental gradient iteration coefficient updates and is divided into three sub-steps:

The first step, calculates the Fourier Transform vector of Current high resolution thing functionCorresponding cost function value ε^kCorresponding Error function:

${\mathrm{\&epsiv;}}^k=\underset{i}{\&Sigma;}\left|\right|\sqrt{I_i}-\left|F^{-1}{P}_i{O}_N^k\right||{|}^2$

Second step, with cost function value ε^kBeing updated incremental gradient iteration coefficient for criterion, more new formula is as follows:

3rd step, outer circulation iterations adds 1, i.e. k ← k+1, interior loop iteration number of times clear 0, i.e. i ← 0.

Image Iterative reconstructing method towards large visual field high resolution micro-imaging the most according to claim 1 and 2, it is special Levying and be in step 5, the implementation process stopping iteration judgement is: as incremental gradient iteration coefficient α^kThe threshold given less than one During value, stop iteration, the Fourier Transform vector of the high-resolution thing function now iteration obtainedCarry out inverse Fourier to become Transformation is changed in spatial domain

$O_{result}=F^{-1}{O}_N^k$

I.e. obtain the distribution of amplitudes of testing sample large visual field high resolution | O_result| and PHASE DISTRIBUTION angle (O_result), its Middle angle () represents and takes argument of complex number；If stop condition is unsatisfactory for, then return step 3, re-start circulation in once, Until meeting stop condition.

Image Iterative reconstructing method towards large visual field high resolution micro-imaging the most according to claim 1 and 2, it is special Levy and be before the image acquisition of step one, it is ensured that the height of object stage (3) is correct, carries out object stage height according to the following steps Regulation:

1. LED array (4) is as microscopical lighting source, lights the most all LED pixel；

2. by the altitude mixture control of object stage (3) to minima z₀, gather corresponding image

Solve the most as followsCorresponding out of focus measure function value

${\mathrm{Laplace}}_{z_0}(x,y)=|2I_{z_0}(x,y)-I_{z_0}(x-step,y)-I_{z_0}(x+step,y)|+|2I_{z_0}(x,y)-I_{z_0}(x,y-step)-I_{z_0}(x,y+step)|$

$L_{z_0}={\mathrm{\&Sigma;\&Sigma;Laplace}}_{z_0}(x,y),{\mathrm{for\&Sigma;\&Sigma;Laplace}}_{z_0}(x,y)>T$

Wherein step is that gradient solves step-length, and T is Grads threshold；

4. the height of object stage (3) is brought up to z1, gather corresponding image

5. according to step 3. in formula, solveCorresponding out of focus measure function

6. repeat said process, until find all in the range of minimum out of focus measure function valueNow corresponding loading The height of platform (3) is i.e. right value, now ensures that microcobjective 1 carries out blur-free imaging to sample 2.

Image Iterative reconstructing method towards large visual field high resolution micro-imaging the most according to claim 1 and 2, its The image sequence selection mode being characterised by iterative reconstruction in step 3 is: LED from the most central optical axis to periphery away from light Axle is sequentially updated；Or press the intensity level of image | | I_i| | descending is updated.