Composite structured light illumination holographic single-exposure imaging method

Abstract

The present application belongs to the field of optics and artificial intelligence technology field, and specifically relates to a kind of composite structure light illumination holographic single exposure imaging method, comprising the following steps: step S1: the 0 phase shift hologram of sample after composite fringe modulation is collected by the Fresnel incoherent self-interference digital holographic system of structured light illumination;Step S2: the 0 phase shift hologram collected in step S1 is input into the first stage deep learning network, and another two phase shift holograms are predicted;Step S3: the 0 phase shift hologram obtained in step S2 and the two predicted phase shift holograms are operated using a three-step phase shift algorithm, the complex amplitude information of the sample is extracted, and the sample is reconstructed by the Fresnel back propagation algorithm to obtain the FINCH reconstruction graph with composite fringe illumination.The present application realizes high-resolution FINCH imaging, and reduces the number of required collected images to one, while maintaining the improvement of lateral resolution, significantly improving the imaging efficiency.

Description

Technical Field

[0001] This invention belongs to the fields of optics and artificial intelligence, and specifically relates to a composite structured light illumination holographic single-exposure imaging method. Background Technology

[0002] Fresnel incoherent correlation holography (FINCH) is a 3D computational imaging technique that utilizes incoherent optical self-interference mechanisms to achieve holographic recording and numerical reconstruction. It uses an incoherent light source to record the 3D information of an object without scanning. FINCH loads a pre-designed lens phase onto a spatial modulator, then uses pixel multiplexing or polarization multiplexing to split the light wave from a single self-emitting point wave or a single point reflection into two spherical light waves with different curvatures. These two spherical light waves then interfere at the camera plane to form a zone-sheet-like coaxial point hologram. The entire object's hologram can be considered as an incoherent superposition of countless point holograms. This hologram records the object's 3D information, which can then be reconstructed using phase-shifting techniques and backpropagation algorithms. Furthermore, FINCH breaks the limitations of Lagrange invariants in traditional imaging, thus achieving superior lateral resolution compared to traditional imaging methods. Due to its advantages such as incoherent illumination, high-resolution imaging, and digital reconstruction, FINCH has shown broad application prospects in fields such as three-dimensional fluorescence microscopy, super-resolution imaging, and color imaging.

[0003] While FINCH overcomes the limitations of Lagrange invariants, resulting in higher lateral resolution, the system's numerical aperture remains unchanged, leaving room for further resolution improvement. Furthermore, FINCH image reconstruction requires at least three holograms to stably remove background and conjugate images of the object, reducing imaging efficiency. To improve the lateral resolution of FINCH, some studies utilize synthetic aperture methods, constructing a larger aperture through multi-view sampling and data stitching. However, this method requires repeated device movement, leading to poor stability. Other research introduces structured illumination into the FINCH system (SI-FINCH), which uses periodic fringes with different directions and phases to shift high-frequency information of the sample to within the system's cutoff frequency, achieving higher lateral resolution than FINCH while avoiding device movement. Although SI-FINCH significantly improves imaging performance, its reconstruction process typically requires multiple illumination directions, multiple phase fringes, and multiple phase-shifted holograms, resulting in a significant increase in the number of holograms the system needs to acquire. To improve acquisition efficiency, research has focused on system improvements, including off-axis interferometry, using polarized cameras, employing dual-channel systems to record holograms, and loading composite structured light illumination patterns onto spatial light modulators (SLMs). However, off-axis configurations and polarized cameras sacrifice the resolution of the reconstructed pattern and reduce the camera's spatial bandwidth product; dual-channel systems often require additional beam splitters, and the positions of the two cameras need to be strictly matched, reducing light energy utilization and increasing system complexity; while composite structured light illumination can reduce the number of acquisitions, it requires constructing a higher-order separation matrix (with three directions) for spectral separation. 2 directions This significantly increases the difficulty of reconstruction. Therefore, how to further reduce the number of acquisition frames while maintaining the high lateral resolution of SI-FINCH remains an important problem to be solved in this field. Summary of the Invention

[0004] The purpose of this invention is to provide a composite structured light illumination holographic single-exposure imaging method, which achieves high-resolution FINCH imaging and reduces the number of images to be acquired to one, thereby significantly improving imaging efficiency while maintaining improved lateral resolution.

[0005] The specific technical solution adopted by this invention is as follows: A composite structured light illumination holographic single-exposure imaging method includes the following steps: Step S1: Acquire the 0-phase-shift hologram of the sample after composite stripe modulation using a Fresnel incoherent self-interference digital holographic system illuminated by structured light; Step S2: Input the 0-phase-shift hologram acquired in step S1 into the first-stage deep learning network to predict the other two phase-shift holograms; Step S3: The three-step phase-shift algorithm is used to process the 0-phase-shift hologram obtained in step S2 and the two predicted phase-shift holograms to extract the complex amplitude information of the sample. Then, the sample is reconstructed using the Fresnel backpropagation algorithm to obtain the FINCH reconstruction image with composite fringe illumination. Step S4: Input the FINCH reconstruction image with composite stripe illumination obtained in step S3 into three independent second-stage deep learning networks to recover the FINCH reconstruction images corresponding to structured light illumination in the 0° and 90° directions under different phase conditions, and obtain six FINCH reconstruction images with cosine stripes of different directions and phase shifts required for SI-FINCH reconstruction. Step S5: The six FINCH reconstructed images obtained in step S4 are processed using the frequency domain structured light reconstruction algorithm to obtain the final high-resolution DL-cSI-FINCH reconstructed image.

[0006] Furthermore, the first-stage deep learning network is a shared encoder-dual decoder U-Net network, which balances the preservation of low-level details and the learning of high-level features through multi-scale feature extraction, skip connections and batch normalization design.

[0007] Furthermore, the second-stage deep learning network is a lightweight shared encoder-dual decoder Mamba U-Net network. The Mamba module in the lightweight shared encoder-dual decoder Mamba U-Net network is used to model long-distance dependencies and global features in the image. The U-Net structure in the lightweight shared encoder-dual decoder Mamba U-Net network preserves local details and edge features through skip connections, realizing the collaborative modeling of global and local information.

[0008] Furthermore, the operations performed on the six FINCH reconstructed images in step 5 include parameter estimation, spectral separation, and spectral stitching.

[0009] A composite structured light illumination holographic single-exposure imaging system includes a structured light illumination modulation unit and a FINCH imaging unit; The structured light illumination modulation unit includes an LED light source, lens A, aperture, lens B, and a digital micromirror device. The light emitted by the LED light source passes through lens A, aperture, and lens B and then illuminates the digital micromirror device. The FINCH imaging unit includes a collimating and shrinking lens group, a third lens, a fourth lens, a first polarizer, a beam splitter, a spatial light modulator, a second polarizer, and a camera unit. The collimating and shrinking lens group, the third lens, the fourth lens, the first polarizer, the beam splitter, and the spatial light modulator are sequentially arranged after the 4f system along the LED light source optical path. The second polarizer and the camera unit are arranged on the reflected optical path of the spatial light modulator after passing through the beam splitter.

[0010] Furthermore, the digital micromirror device employs a time integration method to load periodic and composite stripes.

[0011] Furthermore, the angle between the first polarizer and the second polarizer and the spatial light modulator is 45°.

[0012] An application of a composite structured light illumination holographic single-exposure imaging method is proposed, which applies the composite structured light illumination holographic single-exposure imaging method to three-dimensional fluorescence microscopy, super-resolution imaging, color imaging, or biomedical imaging.

[0013] The technical effects achieved by this invention are as follows: (1) The composite structured light illumination holographic single exposure imaging method of the present invention achieves higher resolution imaging of FINCH by designing a dual-stage deep learning network, and reduces the number of images to be acquired to 1, thereby significantly improving imaging efficiency while maintaining the improvement of horizontal resolution.

[0014] (2) The composite structured light illumination holographic single exposure imaging method of the present invention uses the sample hologram of composite stripe illumination as the input of two-stage deep learning, integrates deep learning prediction with physical models such as angular spectrum propagation and structured light reconstruction, and outputs DL-cSI-FINCH image; it solves the problem of solving the high-order separation matrix and the information shortness of unidirectional structured light, and also takes into account the performance advantages of data-driven methods and the physical rationality of the imaging process. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the SI-FINCH imaging principle of the present invention; Figure 2 This is a schematic diagram of the structure of the composite structured light illumination holographic single-exposure imaging system of the present invention; Figure 3 This invention relates to the DL-cSI-FINCH reconstruction process and network model; Figure 4 This is the output result of the first phase network test set of this invention; Figure 5 This is the output result of the second phase network test set of this invention; Figure 6 This is a comparison of reconstruction results using different imaging methods of the present invention; Figure 7 This is a comparison of the training and reconstruction results of different networks in the second stage of this invention.

[0016] The attached diagram lists the components represented by each number as follows: 1. LED light source; 2. Lens A; 3. Aperture; 4. Lens B; 5. Digital micromirror device; 6. Lens 1; 7. Lens 2; 8. Object; 9. Lens 3; 10. Lens 4; 11. First polarizer; 12. Beam splitter; 13. Spatial light modulator; 14. Second polarizer; 15. Camera. Detailed Implementation

[0017] To make the objectives and advantages of this invention clearer, the invention will be specifically described below with reference to embodiments. It should be understood that the following text is merely used to describe one or more specific embodiments of the invention and does not strictly limit the scope of protection specifically claimed by the invention.

[0018] like Figures 1-7 As shown, a composite structured light illumination holographic single-exposure imaging method is proposed. Based on the FINCH optical path, composite periodic fringe illumination is adopted to composite the 0-phase fringes in the 0° and 90° directions onto a single image. The DMD is loaded and projected onto the object, and a 0-phase-shift hologram of the object is acquired. This hologram is then used to predict all intermediate images for SI-FINCH reconstruction through a two-stage deep learning network. Finally, a high-resolution DL-cSI-FINCH reconstructed image is reconstructed using the predicted intermediate images. This method deeply integrates deep learning prediction with physical imaging models such as angular spectrum propagation and structured light reconstruction algorithms. It not only leverages the advantages of data-driven methods but also ensures the physical rationality of the imaging process. Ultimately, it achieves high-quality reconstruction of DL-cSI-FINCH under single-exposure conditions, solving the problems of low multi-frame acquisition efficiency and insufficient unidirectional structured light illumination information in traditional methods. It can be applied to three-dimensional fluorescence microscopy, super-resolution imaging, color imaging, or biomedical imaging. Specifically, a composite structured light illumination holographic single-exposure imaging method includes the following steps: Step S1: Acquire the 0-phase-shift hologram of sample 8 after composite stripe modulation using a Fresnel incoherent self-interference digital holographic system illuminated by structured light; Step S2: Input the 0-phase-shift hologram acquired in step S1 into the first-stage deep learning network ( Figure 3Model 1 in the model predicts two other phase-shifted holograms, 2π / 3 and 4π / 3. The first-stage deep learning network is a shared encoder-dual decoder U-Net network. Through multi-scale feature extraction, skip connections and batch normalization design, it efficiently balances the preservation of low-level details and the learning of high-level features. It can make full use of the intrinsic relationship between phase-shifted holograms while reducing the number of model parameters, improve the prediction accuracy of missing holograms, and enable the network to accurately predict images. Step S3: The three-step phase shift algorithm is used to process the 0 phase shift hologram obtained in step S2 and the two predicted phase shift holograms to extract the complex amplitude information of sample 8. Then, sample 8 is reconstructed by the angular spectrum propagation algorithm to obtain the FINCH reconstruction image with composite fringe illumination. Step S4: Input the FINCH reconstruction map with composite stripe illumination obtained in step S3 into three independent second-stage deep learning networks respectively. Figure 3 Models 2-4 in the framework recover the FINCH reconstructed images corresponding to structured light illumination at 0° and 90° under different phase conditions, obtaining six FINCH reconstructed images with cosine fringes of different directions and phase shifts required for SI-FINCH reconstruction. The second-stage deep learning network is a lightweight shared encoder-dual decoder Mamba U-Net network. The Mamba module can effectively model long-range dependencies and global features in the image, and is more suitable for processing the spectral distribution of structured light fringes and global texture information in the FINCH reconstructed image than traditional convolutional networks. At the same time, the skip connections of the U-Net structure can preserve local details and edge features, realizing the collaborative modeling of global and local information. This framework compresses the multiple structured light illuminations and multi-step phase shift acquisition process required by traditional SI-FINCH into a single exposure, significantly improving imaging efficiency while maintaining high spatial resolution and structural fidelity. Step S5: Use the frequency domain structured light reconstruction algorithm to perform parameter estimation, spectrum separation and spectrum stitching on the six FINCH reconstructed images obtained in step S4 to obtain the final high-resolution DL-cSI-FINCH reconstructed images.

[0019] The specific process of the composite structured light illumination holographic single-exposure imaging method is as follows: Figure 3 As shown, where, Figure 3 (a) is the DL-cSI-FINCH flowchart, where θ represents the stripe direction, φ represents the stripe phase, and Model 2 to Model 4 represent three different models trained with the same network input and the same network framework. Figure 3 (b) is the first phase of the reconstruction process; Figure 3 (c) represents the second-stage reconstruction process, where I1~I6 represent the second-stage network output. Figure 3 F1~F6 represent the corresponding Fourier transforms; Figure 3(d) represents the network framework for the first phase; Figure 3 (e) is the second-stage network framework.

[0020] Among them, such as Figure 2 As shown, the Fresnel incoherent self-interference digital holographic system with structured light illumination in step S1 includes a structured light illumination modulation unit and a FINCH imaging unit. The structured light illumination modulation unit includes an LED light source 1 (THORLABS M625L4-C1, 630mW, λ=625nm), lens A2, aperture 3, lens B4, and a digital micromirror device 5 (DMD, Fldiscovery F4300, 1920×1080 pixels, pixel size 10.8μm). The light emitted by the LED light source 1 is irradiated onto the digital micromirror device 5 after passing through lens A2, aperture 3, and lens B4. The digital micromirror device 5 uses a time integration method to load periodic stripes and composite stripes. The FINCH imaging unit includes a collimating and shrinking lens group, a third lens 9, a fourth lens 10, a first polarizer 11, a beam splitter 12, a spatial light modulator 13, a second polarizer 14, and a camera unit 15. The collimating and shrinking lens group, the third lens 9, the fourth lens 10, the first polarizer 11, the beam splitter 12, and the spatial light modulator 13 are sequentially arranged after the 4f system along the optical path of the LED light source 1. The second polarizer 14 and the camera unit 15 are arranged on the reflected optical path of the spatial light modulator 13 after passing through the beam splitter 12. The collimating and shrinking lens group includes a 4f system consisting of lens 6 and lens 7. The 4f system is located behind the digital micromirror device 5 and is used to project the periodic and composite fringes reflected by the digital micromirror device 5 onto the surface of the sample 8. At this time, the sample 8 is modulated by the periodic and composite fringes loaded by the DMD. After being modulated by sample 8, the light beam passes sequentially through lens 3 9, lens 4 10 and the first polarizer 11 (45° polarization direction) and then through beam splitter 12 to reach spatial light modulator 13 (Hamamatsu, X15213-16, 1280×1024 pixels, 12.5µm pixel pitch). After being phase-modulated by the lens loaded on spatial light modulator 13, a pair of orthogonally polarized spherical light waves with different curvatures are formed. The two spherical light waves are reflected from the surface of spatial light modulator 13 and then pass through beam splitter 12 to reach the second polarizer 14 (45° polarization direction) for polarization analysis, becoming linearly polarized light with the same polarization direction. Self-interference occurs in the plane of camera unit 15 (LUCID-TRI050S-MC, 4096×3000 pixels, pixel size 3.45μm) to form a 0-phase-shift hologram of sample 8 after being modulated by composite fringes. Finally, the hologram is recorded by camera unit 15, preferably a CMOS camera. Dataset Acquisition and Preprocessing: A total of 2500 datasets were acquired using a Fresnel incoherent self-interfering digital holographic system with structured light illumination. Each dataset contains 27 images (18 holograms and 3 composite fringe modulation holograms for training, and 6 traditional imaging images with structured light modulation for structured light reconstruction and comparison). Each image is 512×512 pixels in size. The 8 loaded samples are images obtained by smoothing and randomly cropping the BioSR dataset. The ratio of training, validation, and test sets for all networks is 8:1:1. All networks use a hybrid loss combining L1 loss and structural similarity loss. Training was performed on a Linux system using an NVIDIA GeForce RTX3090, with 300 training epochs, a batch size of 12, an initial learning rate of 0.0001, and a cosine annealing strategy.

[0021] Network performance testing and verification: The trained network model is applied to an independent test set, and the predicted graphs of each network are compared with the actual acquired phase-shifted images; like Figures 4-5 As shown, Figure 4 The output results for the first phase of the network test set are as follows: Figure 4 (a1) and (a2) are the two outputs of the first-stage network, respectively. Figure 4 (a3) represents the reconstructed images obtained from network input 1 and output 2. Figure 4 (b1)-(b3) are the corresponding labels. Figure 4 (c1)-(c3) represent the normalized intensity contrast of the corresponding row 256 of the image. Figure 4 (d) Figure 4 (e) shows the structural similarity and peak signal-to-noise ratio box plots of the network test set, with the scale bar at 300 μm.

[0022] Figure 5 The output results are for the second-stage network test set. (a1)-(a4), (b1)-(b4), and (c1)-(c4) are the network outputs and corresponding labels for three different models, respectively. The scale bar in the figure is 300μm.

[0023]

[0024] The reconstruction accuracy of the network was evaluated from dimensions such as intensity distribution consistency and structural feature restoration degree, and the model's ability to map different phase features and its generalization performance were verified.

[0025] Comprehensive verification of imaging results: To verify the practical imaging feasibility of the proposed method, this technical solution conducts a quantitative comparative analysis of the reconstruction results of the proposed method with those of wide-field direct imaging, traditional three-step phase-shifting FINCH reconstruction imaging, structured light illumination (SI) reconstruction imaging, and traditional SI-FINCH reconstruction imaging. Structural similarity (SSIM) and peak signal-to-noise ratio (PSNR) are selected as objective evaluation indicators.

[0026] like Figure 6 As shown, Figure 6 To compare the reconstruction results of different imaging methods Figure 6 (a1)-(a5) represent wide-field imaging, FINCH imaging, SI imaging, the method proposed in this technical solution, and SI-FINCH imaging, respectively. Figure 6 (b1)-(b5) and Figure 6 (c1)-(c5) respectively correspond to Figure 6 (a1)-(a5) Enlarged areas within the cyan and yellow dashed boxes; Figure 6 (d) shows the structural similarity and peak signal-to-noise ratio box plots for each image, with the indices calculated based on the SI-FINCH imaging. Figure 6 (e) represents the magnified cyan area along... Figure 6 (b5) Normalized intensity profile of the dashed line; Figure 6 (f) shows the enlarged yellow area. Figure 6 (c5) The result of Gaussian fitting of the normalized intensity distribution of the dashed line; like Figure 7 As shown, Figure 7 This is a comparison of the reconstruction results of different networks in the second stage. Figure 7 (a1)-(a4) are the reconstruction results of ground values, UNet, ResUNet and Mamba UNet predicted values, respectively; Figure 7 (b1)-(b4) are the enlarged areas of the corresponding dashed boxes; Figure 7 (c1)-(c4) are Figure 7 (b1)-(b4) and Figure 7 (b4) error.

[0027] Experimental results show that the proposed SI-FINCH single-exposure reconstruction method can achieve the same resolution and imaging fidelity as the reconstruction results of 18 original images with only one acquisition, which fully verifies the imaging performance and application value of the method.

[0028] Please see Figure 1 As shown, the Figure 1 This is a schematic diagram of the SI-FINCH system (a Fresnel incoherent correlation holographic single-exposure imaging system with structured light illumination). This technical solution incorporates... Figure 1 A detailed explanation of the principles of the SI-FINCH system is provided. For the structured light illumination modulation section: the spatially incoherent light emitted by LED light source 1 is collimated and concentrated by a lens group consisting of lens A2, aperture 3, and lens B4, and then directly incident on the surface of the digital micromirror device 5DMD, which is loaded with a sinusoidal grating stripe pattern. Subsequently, this pattern is formed by lens 6L1 and lens 7L2. The system is scaled down and projected onto the plane of sample 8 to achieve structured light illumination modulation.

[0029] For the FINCH imaging section: the plane of sample 8 is located at the front focal plane of lens 39L3. , d 1 represents the distance between lens 3 (9L3) and lens 4 (10L4). d 2 represents the distance between lens 4 (10L4) and spatial light modulator (13SLM). The distance between the spatial light modulator 13 and the camera unit 15 represents the distance between the first polarizer 11P1 and the second polarizer 14P2 and the SLM. The angle between them is 45°.

[0030] To simplify the analysis, this paper selects a point source in the object plane. The emitted wavelength is Monochromatic incoherent light is used as the incident light wave of the FINCH system. This light wave is collected and collimated by lens 39L3, where the vertical polarization component is focused by a focal length of... The 10L4 lens converges the wave into a spherical wave 1, while the horizontal polarization component is first converged by the 10L4 lens and then focused by the SLM with a focal length of 1. After phase modulation by the lens, it becomes spherical wave 2. These two spherical waves are analyzed by the second polarizer 14P2 and then undergo self-interference in the plane of camera unit 15 (where the two spherical waves have the maximum overlap area), forming a zone plate coaxial point source hologram (PSH). Its expression is: (1) In the formula, the horizontal coordinate and These represent the two-dimensional lateral coordinates of the object plane and the camera unit 15 plane, respectively. For object point Light intensity, It is a complex constant. Represents the imaginary unit, symbol This represents a two-dimensional spatial convolution operation. These represent phase shifts of 0, 2π / 3, and 4π / 3, respectively. This represents the aperture region of a Fresnel zone plate pattern. (Simplified form) and Representing the linear phase function and the quadratic phase function respectively, to eliminate background noise and twin artifacts, this paper uses a three-step phase-shifting algorithm to superimpose three different phase-shifting point source holograms with the form of formula (1) to obtain the complex value PSH, the expression of which is: (2) In the formula It is a complex constant. To reconstruct the planar coordinates, The lateral magnification of the imaging system. This represents the distance from the plane of camera unit 15 to the reconstruction plane.

[0031] The distance of the complex-valued PSH in equation (2) is obtained by integrating the Fresnel diffraction. The reconstructed image obtained by numerical refocusing is the point spread function (PSF) of the FINCH system, which can be expressed as follows according to Fourier optics theory: (3) In the formula Represents a two-dimensional Fourier transform. For scaling operators, make ,in It is a first-order Bessel function of the first kind. FINCH is essentially an incoherent imaging technique; its hologram is formed by the incoherent superposition of the PSH intensities of each object point, thus possessing a cutoff frequency twice that of a coherent system. As can be seen from equation (3), the PSF of the FINCH system exhibits a Bessel function shape similar to that of a coherent system, thus endowing the system with a flat modulation transfer function (MTF) without attenuation. FINCH achieves enhanced system resolution by reshaping the spatial morphology of the MTF, thus balancing high cutoff frequency and high contrast transfer.

[0032] Structured light illumination modulation can extract high spatial frequency information beyond the cutoff frequency of traditional imaging systems. When sample 8 is illuminated by structured light, the reconstructed image of the FINCH system can be represented as follows: (4) in, The light intensity distribution of sample 8 is shown. The spatial frequency vector is The cosine fringe pattern, when the subscript m takes the values ​​1 and 2, corresponds to two cosine fringe patterns with mutually orthogonal directions. It is a constant. Indicates modulation depth. These represent initial phases of 0, 2π / 3, and 4π / 3, respectively. To analyze the super-resolution principle of structured light illumination in the frequency domain, a Fourier transform is performed on equation (4): (5) Symbols in the formula This represents performing a Fourier transform on the original function. and These represent the spatial frequency vectors of the object plane and the reconstruction plane, respectively. It can be seen that under structured light illumination modulation, the spectrum of sample 8 is split into three components: and .in, Corresponding to the sample 8 spectra under traditional uniform illumination, the two newly generated spectra relatively Generate separately The frequency shift. To solve for each frequency spectral component from equation (5), it is necessary to simultaneously establish structured light illumination patterns with phase shifts of 0, 2π / 3, and 4π / 3, and write them in matrix form as follows: (6) The fringe spatial frequency, initial phase, and modulation depth parameters in equation (6) can be estimated with high accuracy using principal component analysis. By inverting its coefficient matrix, the following three spectral components can be separated: (7) After frequency shift matching and splicing of the spectral components calculated from the two orthogonal grating directions, an extended new spectrum can be obtained, expressed as: (8) Finally, by performing an inverse Fourier transform on equation (8), the SI-FINCH reconstructed image can be obtained. .

[0033] The above description is merely a preferred embodiment of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention. Structures, devices, and operating methods not specifically described or explained in this invention are implemented according to conventional methods in the art unless otherwise specified or limited.

Claims

1. A method for composite structured light illumination holographic single-exposure imaging, characterized in that: Includes the following steps: Step S1: Acquire the 0-phase-shift hologram of the sample (8) after composite stripe modulation using a Fresnel incoherent self-interference digital holographic system illuminated by structured light; Step S2: Input the 0-phase-shift hologram acquired in step S1 into the first-stage deep learning network to predict the other two phase-shift holograms; Step S3: The three-step phase shift algorithm is used to calculate the 0 phase shift hologram obtained in step S2 and the two predicted phase shift holograms to extract the complex amplitude information of sample (8). Then, the sample (8) is reconstructed by the angular spectrum propagation algorithm to obtain the FINCH reconstruction image with composite stripe illumination. Step S4: Input the FINCH reconstruction image with composite stripe illumination obtained in step S3 into three independent second-stage deep learning networks to recover the FINCH reconstruction images corresponding to structured light illumination in the 0° and 90° directions under different phase conditions, and obtain six FINCH reconstruction images with cosine stripes of different directions and phase shifts required for SI-FINCH reconstruction. Step S5: The six FINCH reconstructed images obtained in step S4 are processed using the frequency domain structured light reconstruction algorithm to obtain the final high-resolution DL-cSI-FINCH reconstructed image.

2. The composite structured light illumination holographic single-exposure imaging method according to claim 1, characterized in that: The first-stage deep learning network is a shared encoder-dual decoder U-Net network. The shared encoder-dual decoder U-Net network balances the preservation of low-level details and the learning of high-level features through multi-scale feature extraction, skip connections and batch normalization design.

3. The composite structured light illumination holographic single-exposure imaging method according to claim 1, characterized in that: The second-stage deep learning network is a lightweight shared encoder-dual decoder Mamba U-Net network. The Mamba module in the lightweight shared encoder-dual decoder Mamba U-Net network is used to model long-distance dependencies and global features in the image. The U-Net structure in the lightweight shared encoder-dual decoder Mamba U-Net network preserves local details and edge features through skip connections, realizing the collaborative modeling of global and local information.

4. The composite structured light illumination holographic single-exposure imaging method according to claim 1, characterized in that: The operations performed on the six FINCH reconstructed images in step 5 include parameter estimation, spectral separation, and spectral stitching.

5. A composite structured light illumination holographic single-exposure imaging system, which is the Fresnel incoherent self-interference digital holographic system with structured light illumination as described in claim 1, characterized in that: Includes a structured light illumination modulation unit and a FINCH imaging unit; The structured light illumination modulation unit includes an LED light source (1), a lens A (2), an aperture (3), a lens B (4), and a digital micromirror device (5). The light emitted by the LED light source (1) passes through the lens A (2), the aperture (3), and the lens B (4) and then illuminates the digital micromirror device (5). The FINCH imaging unit includes a collimating beam-shrinking lens group, a third lens (9), a fourth lens (10), a first polarizer (11), a beam splitter (12), a spatial light modulator (13), a second polarizer (14), and a camera unit (15). The collimating beam-shrinking lens group, the third lens (9), the fourth lens (10), the first polarizer (11), the beam splitter (12), and the spatial light modulator (13) are arranged sequentially after the 4f system along the optical path of the LED light source (1). The second polarizer (14) and the camera unit (15) are arranged on the reflected optical path of the spatial light modulator (13) after passing through the beam splitter (12).

6. The composite structured light illumination holographic single-exposure imaging system according to claim 5, characterized in that: The digital micromirror device (5) uses a time integration method to load periodic stripes and composite stripes.

7. A composite structured light illumination holographic single-exposure imaging system according to claim 5, characterized in that: The collimating and shrinking lens group includes lens one (6) and lens two (7).

8. A composite structured light illumination holographic single-exposure imaging system according to claim 5, characterized in that: The angle between the first polarizer (11) and the second polarizer (14) and the spatial light modulator (13) is 45°.

9. An application of a composite structured light illumination holographic single-exposure imaging method, characterized in that: The composite structured light illumination holographic single-exposure imaging method according to any one of claims 1-4 is applied to three-dimensional fluorescence microscopy, super-resolution imaging, color imaging, or biomedical imaging.

Images