A lensless quantitative phase imaging and super-resolution method based on information undersampling

By using self-focusing algorithm and noise discrimination matrix to separate signals and noise in lensless microscopy technology, the technical problem of sensor noise affecting phase reconstruction is solved, and more efficient quantitative phase reconstruction and imaging quality is achieved.

CN118587094BActive Publication Date: 2024-12-03NANJING UNIV OF SCI & TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411035009.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-31
Publication Date
2024-12-03
Estimated Expiration
2044-07-31

AI Technical Summary

Technical Problem

In lensless microscopy technology, the noise introduced by the sensor is difficult to be effectively robustly constrained, affecting the accuracy and imaging quality of the phase reconstruction process.

Method used

Using lensless quantitative phase imaging and super-resolution method based on information undersampling, the defocus distance is determined through the self-focusing algorithm, the complex amplitude guess of the focus plane is constructed, and the noise discrimination matrix and the adaptive threshold are used to generate an update coefficient matrix to gradually separate the signal and Noise, quantitative phase reconstruction is achieved.

Benefits of technology

Without knowing the noise statistics in advance, the separation effect between signal and noise is significantly improved, and the accuracy of phase recovery and imaging quality are improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118587094B_ABST
    Figure CN118587094B_ABST
Patent Text Reader

Abstract

The present invention discloses a lensless quantitative phase imaging and super-resolution method based on information undersampling, introduces a pixel discrimination factor to separate noise and signal, constructs an undersampling update coefficient matrix, and provides an adaptive intensity update scheme for phase reconstruction. Based on a traditional lensless on-chip microscope, the present invention provides a new solution to remove the influence of the inevitable noise in the recording process on the reconstruction result, improves the reconstruction accuracy and robustness under the premise of information undersampling, and reduces the dependence of the experimental system on device performance.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to lensless optical microscopic imaging technology, in particular to a lensless quantitative phase imaging and super-resolution method based on information undersampling. Background Art

[0002] Lensless microscopy is an innovative microscopy method that does not rely on traditional optical lens systems, but uses computational imaging technology to achieve high-resolution and large-field imaging. This technology has developed rapidly in the past decade, especially the Aydogan Ozcan team has made significant contributions to improving the resolution of lensless microscopy (Greenbaum, Alon, Wei Luo, Ting-Wei Su, Zoltán Göröcs, Liang Xue, Serhan O. Isikman, Ahmet F. Coskun, Onur Mudanyali, and Aydogan Ozcan. "Imaging without lenses: achievements and remaining challenges of wide-field on-chip microscopy." Nature methods 9, no. 9 (2012): 889-895.). The method they developed not only improved the imaging resolution, but also integrated the super-resolution process and reconstruction process into a unified mathematical framework (Luo, Wei, Yibo Zhang, Zoltán Göröcs, Alborz Feizi, and Aydogan Ozcan. "Propagationphasor approach for holographic image reconstruction." Scientific reports 6, no. 1 (2016): 22738.).

[0003] In lensless microscopy, the pixel size of the imaging sensor is a key factor limiting the reconstruction resolution. Due to the limitation of the manufacturing process, the pixel size cannot be infinitely reduced, so the imaging resolution of lensless microscopy is usually lower than the Nyquist sampling resolution determined by the camera pixel size. To overcome this limitation, the traditional method uses a lateral mechanical sub-pixel displacement of the sample or light source, and processes a series of low-resolution holograms through a pixel super-resolution algorithm to obtain a higher resolution image (Greenbaum, A. and Ozcan, A., 2012. Maskless imaging of dense samples using pixel super-resolution based multi-height lensfree on-chip microscopy. Optics express, 20(3), pp.3129-3143.).

[0004] In recent years, in order to further improve the resolution, researchers have also developed methods such as multi-angle illumination and axial scanning to achieve intensity pixel super-resolution (Greenbaum, A. and Ozcan, A., 2012. Maskless imaging of dense samples using pixel super-resolution based multi-heightlensfree on-chip microscopy. Optics express, 20(3), pp.3129-3143.). Based on these pixel super-resolution technologies, lensless microscopy is expected to become a large-field, high-resolution, low-cost imaging method, providing a long-term living cell observation method for imaging in the biomedical field, especially in the cell culture process.

[0005] The development of lensless microscopy technology has provided a new perspective for biomedical research and clinical diagnosis. Its future development direction will focus on improving imaging quality, reducing system costs, and improving operational convenience. With the improvement of computing power and the advancement of algorithms, lensless microscopy technology is expected to be more widely used in the future. However, lensless imaging is extremely dependent on the imaging quality of the sensor. How to avoid the noise introduced by the sensor and provide a robust constraint method for the phase reconstruction process are problems that need to be solved in lensless imaging. Summary of the invention

[0006] The object of the present invention is to provide a lens-free quantitative phase imaging and super-resolution method based on information undersampling.

[0007] The technical solution to achieve the purpose of the present invention is: a lensless quantitative phase imaging and super-resolution method based on information undersampling, comprising:

[0008] Step 1, collecting initial light intensity images of samples under different conditions;

[0009] Step 2, determine the defocus distance through the autofocus algorithm, and construct the initial complex amplitude of the object under test on the focusing plane as the complex amplitude guess of the focusing plane;

[0010] Step 3, propagate the initial complex amplitude of the focus plane complex amplitude guessed at the focus plane to the sensor surface, and define a noise discrimination matrix according to the difference between the sensor surface intensity map and the intensity map collected by the sensor;

[0011] Step 4, generating an undersampling update coefficient matrix based on the adaptive threshold;

[0012] Step 5: Intensity constraints are imposed on the sensor surface intensity maps whose update coefficients in the update coefficient matrix are 1, and the update coefficients whose values ​​are 0 remain unchanged;

[0013] Step 6, back-propagating the updated sensor surface complex amplitude to the focusing plane to obtain the updated focusing plane complex amplitude, and repeating steps 3 to 6 with the updated focusing plane complex amplitude as the focusing plane complex amplitude guess until convergence, and obtaining the quantitative phase reconstruction of the sample;

[0014] Step 7: Convert the complex amplitude information of the focal plane finally obtained in step 6 into the complex amplitude guess of the focal plane of the next intensity image, repeat steps 3 to 6 for the next intensity image until all images are processed, and obtain the super-resolution quantitative phase reconstruction result.

[0015] Preferably, a lensless on-chip microscopy system is used to collect initial light intensity images of multiple frames of samples, wherein the lensless on-chip microscopy system comprises a light source and a sensor, the sample is arranged at the sensor, and the light source, the sample and the sensor are coaxial.

[0016] Preferably, initial light intensity images of multiple frames of samples are collected at different illumination wavelengths, different illumination angles and light sources, and different lateral displacements of the sample and the sensor.

[0017] Preferably, the specific method for determining the defocus distance by the autofocus algorithm is:

[0018] The initial light intensity images of multiple frames of samples collected under different conditions are sequentially upsampled to construct the complex amplitude of the sensor surface;

[0019] The complex amplitude of the sensor surface is propagated in reverse for different distances, and the evaluation function of the self-focusing is constructed by calculating the gradient sum of the amplitude part of the complex amplitude after propagation for different distances;

[0020] If the sample is an intensity object, it is considered to be focused when the evaluation function value reaches the maximum value; if the sample is a phase object, it is considered to be focused when the evaluation function value reaches the minimum value; the propagation distance corresponding to the focusing is taken as the defocus distance.

[0021] Preferably, the evaluation function of self-focusing is specifically:

[0022]

[0023] in, and Represents the gradient of the amplitude part of the image along the horizontal direction x and the vertical direction y, is the evaluation function of self-focus.

[0024] Preferably, the specific method for constructing the initialization complex amplitude of the object to be measured on the focusing plane is: reversely propagate the sensor surface complex amplitude corresponding to the first intensity image of the collected sample to the defocus distance, obtain the high-resolution initialization complex amplitude of the sample to be measured on the focusing plane, and use it as the complex amplitude guess of the focusing plane.

[0025] Preferably, the noise discriminant matrix is ​​composed of pixel discriminant factors The pixel discrimination factor Pixel The absolute value of the difference between the intensity value calculated at and the intensity value collected by the sensor.

[0026] Preferably, the specific method for generating the undersampling update coefficient matrix based on the adaptive threshold is:

[0027] The elements in the noise discriminant matrix whose values ​​exceed the threshold are reset to 0, and those that do not exceed the threshold are reset to 1;

[0028] The threshold decreases as the number of iterations increases.

[0029] Preferably, the specific formula for performing intensity constraint on the sensor surface intensity map with an update coefficient of 1 in the update coefficient matrix is:

[0030]

[0031] in, is an imaginary unit, is a function of the complex amplitude angle, is the complex amplitude of the sensor surface after intensity constraint, is the update coefficient matrix, is the intensity value collected by the sensor.

[0032] Preferably, the specific method of converting the complex amplitude information of the focal plane finally obtained in step 6 into the complex amplitude guess of the focal plane of the next intensity map is:

[0033] For images collected at different illumination wavelengths, Convert to the next wavelength, the conversion formula is as follows

[0034]

[0035] In the formula, is the complex amplitude information of the focus plane updated in step 6, is the wavelength corresponding to the jth image, is the wavelength corresponding to the j+1th image, is the complex amplitude guess of the focal plane corresponding to the j+1th image;

[0036] For images captured at different lighting angles, it is necessary to convert to the next lighting angle:

[0037]

[0038] is the wavefront modulation corresponding to the j-th illumination angle, is the wavefront modulation corresponding to the j+1th illumination angle;

[0039] For images collected with different lateral displacements of light source, sample and sensor, Translate the horizontal position of the j+1th image:

[0040]

[0041] Represents the horizontal image shift operation.

[0042] Compared with the prior art, the present invention has the following significant advantages:

[0043] The present invention realizes the separation of signal and noise in the phase recovery process by using the difference between the collected original light intensity map and the calculated intensity map without knowing the noise statistics in advance, and the recovery effect is obviously better than the existing method.

[0044] The present invention is further described in detail below in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a schematic diagram of the undersampling update coefficient matrix.

[0046] Figure 2 It is a flow chart of the lensless quantitative phase reconstruction algorithm based on information undersampling.

[0047] Figure 3 Schematic diagram of the experimental setup for multi-wavelength tomography based on lensless on-chip microscopy.

[0048] Figure 4 It is based on the quantitative phase reconstruction and long-term living cell observation results of color sensors. DETAILED DESCRIPTION

[0049] like Figure 2 As shown, a lensless quantitative phase imaging and super-resolution method based on information undersampling includes the following steps:

[0050] Step 1: Build a lensless on-chip microscopy system. In the lensless on-chip microscopy system, the sample is placed close to the sensor so that the light source, sample and sensor are coaxial, and then the original intensity map of the sample is collected. In the image directly collected by the sensor in the lensless on-chip holographic microscope, there are inevitable pixel values ​​that are unavailable due to noise, etc., resulting in undersampling of information. It is necessary to introduce a pixel discrimination factor to separate noise and signal.

[0051] To achieve pixel super-resolution, the lensless quantitative phase recovery algorithm based on information undersampling can be combined with existing lensless super-resolution modulation methods, such as lateral displacement of light source, sample and sensor, multi-wavelength scanning, multi-illumination angle scanning, etc. In the specific implementation, by changing different illumination wavelengths, different illumination angles, different lateral displacements of light source, sample and sensor, the initial light intensity images under different conditions are collected. , is the number of collected initial light intensity images. In the implementation process, one condition can be changed in sequence, and multiple images corresponding to the change of the same condition can be collected, and then another condition can be changed, and multiple images corresponding to the change of the same condition can be collected.

[0052] Step 2: Determine the defocus distance through the autofocus algorithm and construct the initial complex amplitude of the object under test on the focusing plane. The specific process is as follows:

[0053] The initial light intensity images of multiple frames of samples collected under different conditions are upsampled in turn to construct the complex amplitude of the sensor surface. The complex amplitude of the sensor surface is then propagated back to different distances. The evaluation function of self-focusing is constructed by calculating the gradient sum of the amplitude part of the complex amplitude at different distances:

[0054]

[0055] in, and Represents the gradient of the image along the horizontal and vertical directions.

[0056] For intensity objects, when the evaluation function value reaches the maximum value, it is considered to be focused; for phase objects, when the evaluation function value reaches the minimum value, it is considered to be focused; the corresponding propagation distance when focused is the defocus distance .

[0057] The complex amplitude of the sensor surface is propagated back to the focusing plane to obtain the high-resolution initialization complex amplitude of the sample to be tested on the focusing plane , and used as a guess for the complex amplitude of the focal plane .

[0058] Step 3: Predict the complex amplitude of the focal plane Propagate to the sensor surface, and define the noise discrimination matrix by the difference between the calculated intensity map and the original intensity map;

[0059] Lensless on-chip holographic microscopy usually uses coherent or quasi-coherent illumination. The difference in light intensity response of sensor pixels under monochromatic illumination will introduce noise. The dark field noise and bad pixels of the sensor itself, the information loss introduced by overexposure of bright rings in the diffraction pattern, and the pixel value loss caused by the filter layer of the color camera will all cause some pixel values ​​in the image directly collected by the sensor to be wrong, that is, the hologram information is undersampled.

[0060] Predict the complex amplitude of the focal plane Propagate to the sensor surface and obtain the complex amplitude of the sensor surface , and its intensity is recorded as Then compare it with the intensity collected by the sensor, that is, construct the noise discrimination matrix ,matrix The elements of , is the pixel discrimination factor, indicating the pixel The difference between the intensity value calculated at and the intensity value collected by the sensor, is the image intensity collected by the sensor. The further away from 0, the more serious the noise impact or the more overexposed or damaged the image is.

[0061] Step 4, generate an undersampling update coefficient matrix: Based on the adaptive strategy, dynamically adjust the discrimination threshold, set the values ​​exceeding the threshold in the discrimination matrix to 0, and set the values ​​not exceeding the threshold to 1, thereby obtaining the undersampling update coefficient matrix.

[0062] The specific implementation process is: based on the adaptive strategy, dynamically adjust the discrimination threshold, set the values ​​exceeding the threshold in the discrimination matrix to 0, and set the values ​​not exceeding the threshold to 1, so as to obtain the undersampling update coefficient matrix , a typical undersampling update coefficient matrix is Figure 1 As shown in Figure 2. Specifically, the adaptive strategy is that when the iteration (j=1) is just started, the threshold is set to 0.5. As the number of iterations increases, the threshold gradually decreases toward 0. The schematic diagram of the undersampling update coefficient matrix corresponding to the random noise is shown in Figure 2. Figure 1As shown in (a), only the green channel data is used. The red and blue channels can be understood as fixed pattern noise. The corresponding undersampling update coefficient matrix diagram is shown in Figure 1 As shown in (b) in .

[0063] Step 5, sensor surface complex amplitude update: strength constraints are applied to update coefficients of 1, while update coefficients of 0 keep the calculated values ​​unchanged.

[0064] The specific implementation process is: the intensity constraint is applied to the update coefficient of 1, while the calculated value is kept unchanged for the update coefficient of 0, and the phase part remains unchanged. The specific formula for the intensity constraint is:

[0065] ,

[0066] in, is an imaginary unit, It is a function of the complex amplitude angle.

[0067] Step 6: Update the complex amplitude Back propagates to the focal plane to obtain the updated complex amplitude of the focal plane, and repeats steps 3 to 6 with the updated complex amplitude of the focal plane as the guess of the complex amplitude of the focal plane until convergence, and obtains the quantitative phase reconstruction of the sample;

[0068] The specific implementation process is: the updated complex amplitude Then the focus plane is transmitted back to obtain the updated focus plane complex amplitude, the formula is as follows:

[0069] .

[0070] The updated complex amplitude of the focal plane is used as the new complex amplitude guess of the focal plane. Repeat steps 3-6.

[0071] Step 7, convert the complex amplitude information of the focal plane finally obtained in step 6 into the complex amplitude guess of the focal plane of the next intensity image, and repeat steps 3 to 6.

[0072] The specific implementation process is: for images collected at different illumination wavelengths, Convert to the next wavelength, the conversion formula is as follows

[0073]

[0074] In the formula, is the complex amplitude information of the focus plane updated in step 6, is the wavelength corresponding to the jth image, is the wavelength corresponding to the j+1th image.

[0075] For images captured at different lighting angles, it is necessary to convert to the next lighting angle:

[0076]

[0077] is the wavefront modulation corresponding to the j-th illumination angle, is the wavefront modulation corresponding to the j+1th illumination angle.

[0078] For images collected with different lateral displacements of light source, sample and sensor, Translate the horizontal position of the j+1th image:

[0079]

[0080] Represents the horizontal image shift operation.

[0081] During the iteration process, the intensity maps with different diffraction changes will continuously constrain and optimize the reconstruction results, thereby increasing the amount of information and obtaining super-resolution results.

[0082] Example

[0083] A typical application of a lensless quantitative phase imaging and super-resolution method based on information undersampling

[0084] The lensless system based on the color sensor is combined with the multi-wavelength scanning system. The color sensor filter usually adopts Bayer coding (1 red, 1 blue, 2 green), and the wavelength scanning range is set within the range where the green channel has a strong response. When scanning with the green light band, the grayscale value of the red and blue channel pixels is always very low. Compared with the green channel data, it can be compared to noise. Based on the pixel discrimination factor, the pixels with low signal-to-noise ratio are excluded and not updated, that is, the red and blue channel pixels are excluded. The specific steps are as follows:

[0085] Step 1: Build a multi-wavelength illumination lensless on-chip microscopy imaging system to collect the original intensity map of the sample at the corresponding wavelength.

[0086] like Figure 3 As shown, the present invention is based on a conventional lensless on-chip microscopy system, and the lensless on-chip microscopy system comprises: a wavelength scanning light source 1, a sample 2 and a sensor 3. The wavelength scanning light source used in the present invention is realized by combining an acousto-optic tunable filter with a supercontinuum laser, and wavelength scanning can also be realized by coupling multiple monochromatic light sources (lasers, LEDs) or replacing the supercontinuum laser-acousto-optic filter combination with a wavelength scanning laser. Sample 2 is directly placed on the image sensor for imaging, and the present invention uses a board-level monochrome CMOS sensor or a color sensor for image capture.

[0087] The specific implementation process is: use a coherent light source with a wide range of tunable wavelengths, tune the illumination wavelengths in sequence, illuminate the sample, and synchronously trigger the camera to record holographic image sequences at different wavelengths. .

[0088] Step 2: Determine the defocus distance through the autofocus algorithm and construct the initial complex amplitude of the object under test on the focus plane. The specific implementation process is: for the image collected by the color sensor, first use the data of the green channel to interpolate the data of other channels, and then construct the sensor surface complex amplitude based on this. Then propagate the sensor surface complex amplitude in reverse for different distances, and construct the self-focus evaluation function by calculating the gradient sum of the amplitude part of the complex amplitude in this process:

[0089] ,

[0090] in, and Represents the gradient of the image in the horizontal and vertical directions. For intensity objects, when the evaluation function value reaches the maximum value, it is considered to be focused; for phase objects, when the evaluation function value reaches the minimum value, it is considered to be focused, and the corresponding propagation distance at this moment is the self-focusing distance .

[0091] The intensity images collected at different wavelengths are upsampled in turn to construct complex amplitudes, which are back-propagated to the focal plane and averaged to obtain the high-resolution initialization complex amplitude of the sample to be tested on the focal plane. , and used as a guess for the complex amplitude of the focal plane .

[0092] Step 3: Predict the complex amplitude of the focal plane Propagate to the sensor surface, and define the noise discrimination matrix by the difference between the calculated intensity map and the original intensity map;

[0093] Lensless on-chip holographic microscopy usually uses coherent or quasi-coherent illumination. The difference in light intensity response of sensor pixels under monochromatic illumination will introduce noise. The dark field noise and bad pixels of the sensor itself, the information loss introduced by overexposure of bright rings in the diffraction pattern, and the pixel value loss caused by the filter layer of the color camera will all cause some pixel values ​​in the image directly collected by the sensor to be wrong, that is, the hologram information is undersampled.

[0094] Predict the complex amplitude of the focal plane Propagate to the sensor surface and obtain the complex amplitude of the sensor surface , and its intensity is recorded as Then compare it with the intensity collected by the sensor, that is, construct the noise discrimination matrix ,matrix The elements of , is the pixel discrimination factor, indicating the pixel The difference between the calculated intensity value and the collected intensity value, is the image intensity collected by the sensor. The further away from 0, the more serious the noise impact or the more overexposed or damaged the image is.

[0095] Step 4, generate an undersampling update coefficient matrix: Based on the adaptive strategy, dynamically adjust the discrimination threshold, set the values ​​exceeding the threshold in the discrimination matrix to 0, and set the values ​​not exceeding the threshold to 1, thereby obtaining the undersampling update coefficient matrix.

[0096] The specific implementation process is: based on the adaptive strategy, dynamically adjust the discrimination threshold, set the values ​​exceeding the threshold in the discrimination matrix to 0, and set the values ​​not exceeding the threshold to 1, so as to obtain the undersampling update coefficient matrix When scanning the green light band, due to the existence of the Bayer filter layer, the green pixels respond normally, while the red and blue channel pixels respond very low, and the corresponding update coefficient matrix elements are set to 0. The abandoned red and blue channels can be understood as fixed pattern noise with uniform periodic distribution. The corresponding undersampling update coefficient matrix diagram is as follows: Figure 1 As shown in (b) in .

[0097] Step 5: In the sensor plane, use the collected intensity map to constrain the green pixel position in the complex amplitude. The specific implementation process is: use the illumination wavelength The intensity map collected Complex amplitude The updated sensor surface complex amplitude is written as

[0098]

[0099] Step 6: The updated complex amplitude is transmitted back to the focusing plane to update the complex amplitude information of the sample plane. Then the focusing plane is transmitted back to obtain the updated complex amplitude information of the focusing plane. The formula is as follows:

[0100] .

[0101] Step 7: Convert the complex amplitude information of the focal plane obtained in step 6 into the complex amplitude guess of the focal plane of the next intensity image. The conversion formula is as follows:

[0102]

[0103] In the formula, is the complex amplitude information of the focus plane updated in step 6, is the wavelength corresponding to the jth image, is the wavelength corresponding to the j+1th image. Repeat steps 3 to 6 for the next intensity image until all images are processed, and finally obtain the quantitative phase reconstruction of pixel super resolution.

[0104] The present invention introduces a pixel discrimination factor to separate noise and signal for a single-frame hologram containing noise, constructs an undersampling update coefficient matrix, and provides an adaptive intensity update scheme for phase reconstruction. It can be combined with existing pixel super-resolution methods to achieve pixel super-resolution quantitative phase reconstruction, such as Figure 4 As shown on the left. The present invention is aimed at a lensless system based on a color sensor combined with a multi-wavelength scanning phase pixel super-resolution method. It has only one light source at a fixed position, can maintain a relatively high coherence, and does not introduce mechanical displacement, thereby improving stability. It is suitable for long-term living cell observation, such as Figure 4 Shown right.

Claims

1. A lensless quantitative phase imaging and super-resolution method based on information undersampling, characterized in that: include: Step 1, collecting initial light intensity images of samples under different conditions; Step 2, determine the defocus distance through the autofocus algorithm, and construct the initial complex amplitude of the object under test on the focusing plane as the complex amplitude guess of the focusing plane; Step 3: propagate the initial complex amplitude of the focus plane guessed at the focus plane to the sensor surface, and define the noise discrimination matrix according to the difference between the intensity map of the sensor surface and the intensity map collected by the sensor. The noise discrimination matrix consists of the pixel discrimination factors corresponding to the jth image. The pixel discrimination factor Pixel The absolute value of the difference between the intensity value calculated at and the intensity value collected by the sensor; Step 4: Generate an undersampling update coefficient matrix based on the adaptive threshold. The specific method is: The elements in the noise discriminant matrix whose values ​​exceed the threshold are reset to 0, and those that do not exceed the threshold are reset to 1; The threshold value decreases as the number of iterations increases; Step 5: Perform intensity constraints on the sensor surface intensity map with an update coefficient of 1 in the update coefficient matrix, and keep the update coefficient of 0 unchanged. The specific formula for intensity constraints is: ; in, is an imaginary unit, is a function of the complex amplitude angle, is the complex amplitude of the sensor surface after intensity constraint, is the update coefficient matrix, is the intensity value collected by the sensor, is the complex amplitude of the sensor surface; Step 6, back-propagating the updated sensor surface complex amplitude to the focusing plane to obtain the updated focusing plane complex amplitude, taking the updated focusing plane complex amplitude as the focusing plane complex amplitude guess, repeating steps 3 to 6 until convergence, and obtaining the quantitative phase reconstruction of the sample; Step 7: Convert the complex amplitude information of the focal plane finally obtained in step 6 into the complex amplitude guess of the focal plane of the next intensity image, repeat steps 3 to 6 for the next intensity image until all images are processed, and obtain the super-resolution quantitative phase reconstruction result.

2. The lensless quantitative phase imaging and super-resolution method based on information undersampling according to claim 1, characterized in that: The initial light intensity map of multiple frames of samples is collected by using a lensless on-chip microscopy system, wherein the lensless on-chip microscopy system comprises a light source and a sensor, the sample is arranged at the sensor, and the light source, the sample and the sensor are coaxial.

3. The lensless quantitative phase imaging and super-resolution method based on information undersampling according to claim 1, characterized in that: Collect initial light intensity images of multiple frames of samples with different illumination wavelengths, different illumination angles and light sources, different samples, and different lateral displacements of the sensor.

4. The lensless quantitative phase imaging and super-resolution method based on information undersampling according to claim 1, characterized in that: The specific method for determining the defocus distance through the autofocus algorithm is: The initial light intensity images of multiple frames of samples collected under different conditions are sequentially upsampled to construct the complex amplitude of the sensor surface; The complex amplitude of the sensor surface is propagated in reverse for different distances, and the evaluation function of the self-focusing is constructed by calculating the gradient sum of the amplitude part of the complex amplitude after propagation for different distances; If the sample is an intensity object, it is considered to be focused when the evaluation function value reaches the maximum value; if the sample is a phase object, it is considered to be focused when the evaluation function value reaches the minimum value; the propagation distance corresponding to the focusing is taken as the defocus distance.

5. The lensless quantitative phase imaging and super-resolution method based on information undersampling according to claim 4, characterized in that: The evaluation function of self-focusing is specifically: ; in, and Represents the gradient of the amplitude part of the image along the horizontal direction x and the vertical direction y, is the evaluation function of self-focus.

6. The lensless quantitative phase imaging and super-resolution method based on information undersampling according to claim 1, characterized in that: The specific method for constructing the initialization complex amplitude of the object to be measured on the focusing plane is: reversely propagate the sensor surface complex amplitude corresponding to the first intensity image of the collected sample to the defocus distance, obtain the high-resolution initialization complex amplitude of the sample to be measured on the focusing plane, and use it as the complex amplitude guess of the focusing plane.

7. The lensless quantitative phase imaging and super-resolution method based on information undersampling according to claim 1, characterized in that: The specific method of converting the complex amplitude information of the focal plane finally obtained in step 6 into the complex amplitude guess of the focal plane of the next intensity map is: For images collected at different illumination wavelengths, Convert to the next wavelength, the conversion formula is as follows ; In the formula, is the complex amplitude information of the focus plane updated in step 6, is the wavelength corresponding to the jth image, is the wavelength corresponding to the j+1th image, is the complex amplitude guess of the focal plane corresponding to the j+1th image; For images captured at different lighting angles, it is necessary to convert to the next lighting angle: ; is the wavefront modulation corresponding to the j-th illumination angle, is the wavefront modulation corresponding to the j+1th illumination angle; For images collected with different lateral displacements of light source, sample and sensor, Translate the horizontal position of the j+1th image: ; Represents the horizontal image shift operation.