Integrated compressive sensing high-speed imaging device and method

Abstract

The application discloses an integrated compressed sensing high-speed imaging device and method. The application can be directly integrated on a commercial microscope platform. After biological fluorescence is collected by the commercial microscope platform, the biological fluorescence is imaged on a binary amplitude spatial light modulator through a polarization beam splitter prism. The spatial light modulator encodes and modulates the biological fluorescence and reflects the biological fluorescence to the polarization beam splitter prism. The biological fluorescence is reflected on the polarization beam splitter prism and sequentially passes through a first relay lens, a first mirror, a second relay lens and a second mirror and is imaged on a camera. The measurement image obtained on the camera needs to be processed through a three-step algorithm processing procedure including compressed sensing calculation, background removal and deconvolution on a computer to obtain a high spatiotemporal resolution and high contrast biological fluorescence image. The application can improve the spatiotemporal resolution of the commercial microscope, and has the advantages of high integration, compact spatial structure and low cost.

Description

Technical Field

[0001] This invention relates to an integrated compressed sensing high-speed imaging device and method, which relates to the field of optical microscopy, and specifically to a high-speed fluorescence microscopy imaging device and method based on a compressed sensing algorithm. Background Technology

[0002] Fluorescence microscopy utilizes fluorescent dyes, fluorescent proteins, and other techniques to label specific molecules or structures in a sample. These fluorescent labels are then excited using lasers of specific wavelengths, and imaged using optical microscopy. Because fluorescence microscopy can label and image different structures separately and offers advantages such as high sensitivity and high resolution, it has become a crucial information acquisition method in fields such as biology, medicine, and materials science. However, with ever-increasing imaging demands, traditional fluorescence microscopy techniques are no longer sufficient to meet researchers' needs in terms of imaging speed and resolution.

[0003] Compressed sensing imaging technology is based on the mathematical assumption of signal sparsity, utilizing coded random sampling and post-processing algorithms to reconstruct images. Compared to traditional imaging methods, it overcomes the Nyquist sampling theorem, enabling accurate image reconstruction with minimal sampling, significantly improving imaging speed, and ultimately exceeding the frame rate limit of cameras. In recent years, researchers have gradually applied compressed sensing imaging technology to various imaging fields, such as the aperture-coded snapshot spectral imaging technique CASSI in hyperspectral imaging and the compressed ultrafast imaging technique CUP in ultrafast imaging.

[0004] To achieve compressed sensing imaging, improvements are generally needed in traditional imaging systems in terms of hardware coding and algorithm reconstruction.

[0005] In terms of hardware coding, researchers generally use digital micromirror devices (DMDs) to encode the imaging beam. DMDs have advantages such as fast switching speed and high contrast. However, since DMDs generally deflect the beam by 12 degrees and generally deflect the beam along the diagonal of the pixel, this is not friendly to the construction of the system optical path and will lead to space constraints in component design, ultimately making the optical system complex and large.

[0006] In terms of algorithmic reconstruction, researchers have utilized various inverse problem-solving frameworks to reconstruct compressed sensing images, such as the two-step iterative convergence algorithm TwIST, the alternating direction multiplier method (ADMM), and the generalized alternating projection method (GAP). Among these, ADMM is widely used due to its flexible structure, its ability to be combined with various priors, and its high reconstruction quality. However, because ADMM does not consider the blurring caused by the point spread function and the low-frequency background present in the image when reconstructing compressed sensing images, the final reconstructed image may suffer from low contrast and low resolution. Summary of the Invention

[0007] To address the problems existing in the background technology, this invention provides an integrated compressed sensing high-speed imaging device and method. This invention utilizes a binarized amplitude-type spatial light modulator to replace the hardware coding scheme of the DMD and a three-step image reconstruction algorithm that integrates compressed sensing solution algorithms with background removal and deconvolution algorithms. This invention constructs a highly integrated compressed sensing high-speed imaging device using a binarized amplitude-type spatial light modulator. This device occupies little space and can be directly combined with various commercial microscope platforms, providing a low-cost, high-speed acquisition solution for fluorescence microscopy. Simultaneously, with the help of the three-step image reconstruction algorithm, this invention ultimately provides high-speed, low-background, and high-resolution imaging results for fluorescence microscopy.

[0008] An integrated compressed sensing high-speed imaging device includes: The coding and modulation module is used to encode and modulate biological fluorescence and transmit it to the next module; A relay imaging module is used to re-image the modulated biofluorescence onto the camera's target surface. The decoding and reconstruction module is used to convert a single image measured by the camera into multiple high-resolution images.

[0009] Furthermore, the encoding and modulation module includes a polarizing beam splitter and a binarized amplitude spatial light modulator. The direction of fluorescence propagation from the commercial microscope platform is used as the emission direction, and the polarizing beam splitter and the binarized amplitude spatial light modulator are arranged alternately in sequence. The fluorescence from the commercial microscope platform first propagates to the polarizing beam splitter and is split into P-light and S-light of equal intensity. The P-light continues to propagate through the polarizing beam splitter to the binarized amplitude spatial light modulator. This part of the fluorescence is encoded and modulated by the binarized amplitude spatial light modulator, converted into S-light, and reflected back to the polarizing beam splitter. Then, the polarizing beam splitter reflects this part of the fluorescence to the relay imaging module.

[0010] Furthermore, the relay imaging module includes a first relay lens, a first reflector, a second relay lens, a second reflector, and a camera. The first relay lens is perpendicular to the fluorescence direction emitted from the coded imaging module. The first reflector is positioned after the first relay lens, rotating the fluorescence direction counterclockwise by 90 degrees. The second relay lens is perpendicular to the fluorescence propagation direction after the first reflector. The second reflector is positioned after the second relay lens, rotating the fluorescence direction counterclockwise by 90 degrees again. The camera is positioned perpendicular to the fluorescence propagation direction after the second reflector. The fluorescence emitted from the coded modulation module is first deflected by the first relay lens, then rotated counterclockwise by 90 degrees by the first reflector, then deflected by the second relay lens, and then rotated counterclockwise by 90 degrees by the second reflector, finally forming an image on the target surface of the camera.

[0011] Furthermore, the encoding modulation module uses a binary amplitude-type spatial light modulator for encoding modulation of fluorescence. Its 180-degree reflectivity allows the incident and outgoing light paths to coincide, resulting in a shorter optical path length and a more compact structure. The relay imaging module uses two mirrors to further fold and compress the optical path, reducing the overall size of the device. The device is compatible with commercial microscope platforms, receiving fluorescence collected from them, without requiring its own fluorescence collection devices such as microscope objectives.

[0012] Furthermore, the decoding and reconstruction module includes a compressed sensing solution unit, a background removal unit, and a deconvolution unit; The input to the compressed sensing solution unit is an coded mask pattern loaded on a binary amplitude spatial light modulator and a measurement image from the camera, and the output is multiple low-resolution fluorescence images. The input to the background removal unit is multiple low-resolution fluorescence images output by the compressed sensing solution unit, and the output is multiple low-resolution fluorescence images without background. The input to the deconvolution unit is multiple low-resolution fluorescence images without background and a system point spread function image output by the background removal unit, and the output is multiple high-resolution fluorescence images without background. The system point spread function is a response function describing the optical imaging system to imaging of an ideal point light source. The system point spread function image is obtained by imaging the fluorescent microspheres during the system calibration stage. A measurement image obtained by the camera is sequentially processed by a compressed sensing solution unit, a background removal unit, and a deconvolution unit to obtain multiple high-resolution fluorescence images without background.

[0013] Furthermore, the present invention also provides an integrated compressed sensing high-speed imaging method, specifically including: Biological fluorescence emitted from a commercial microscope platform first passes through a polarizing beam splitter. P-light propagates through the beam splitter to a binarized amplitude spatial light modulator (ASL). The ASL encodes and modulates the fluorescence, converting the P-light into S-light. The S-light is then reflected back to the beam splitter, which reflects it to a first relay lens. The fluorescence passing through the first relay lens is reflected by a first mirror to a second relay lens, and then by a second mirror to the camera, completing the imaging process. The measurement image obtained from the camera and the coded mask pattern on the ASL are input into a compressed sensing unit to obtain multiple low-resolution fluorescence images with background. These images are then input into a background removal unit to obtain multiple low-resolution fluorescence images without background. Finally, these images, along with the system point spread function image, are input into a deconvolution unit to obtain multiple high-resolution fluorescence images without background.

[0014] Furthermore, the binarized amplitude spatial light modulator can load any binarized image; at the position where the pixel value is 1, the P light is converted into the S light, reflected to the polarizing beam splitter and then reflected into the first relay lens; at the position where the pixel value is 0, the P light is not converted into the S light, and after being reflected, it passes through the polarizing beam splitter and finally leaves the system.

[0015] Furthermore, the binarized amplitude spatial light modulator is positioned at the biofluorescence image plane of the commercial microscope platform; the first relay lens and the second relay lens form a 4F system, and its front focal point coincides with that of the binarized amplitude spatial light modulator; the camera is positioned at the rear focal point of the 4F system, so the biofluorescence is imaged precisely on the camera, and the binarized amplitude spatial light modulator and the camera form a conjugate plane.

[0016] Furthermore, the aforementioned binarized amplitude spatial light modulator needs to be precisely synchronized with the camera. Depending on the synchronization method, the two can receive synchronization signals from a unified signal source, or the binarized amplitude spatial light modulator can emit a synchronization signal and the camera receives the synchronization signal, or the camera can emit a synchronization signal and the binarized amplitude spatial light modulator receives the synchronization signal. When the camera receives the signal and starts exposure or starts exposure autonomously, the binarized amplitude spatial light modulator begins to load the mask. According to the pre-set control program, the binarized amplitude spatial light modulator changes the mask multiple times during one exposure of the camera.

[0017] Furthermore, the compressed sensing solution unit is composed of various inverse problem solution frameworks and various image priors. The inverse problem solution frameworks include, but are not limited to, two-step iterative convergence algorithms, alternating direction multiplier methods, and generalized alternating projection methods. The various image priors include, but are not limited to, total variational priors, L1 norm priors, Hessian continuity priors, and weighted kernel norm minimization priors. The background removal unit consists of various image background removal algorithms, including, but not limited to, background removal algorithms based on wavelet transform and background removal algorithms based on frequency domain filtering. The deconvolution unit consists of various algorithms that improve the resolution of the original image by using the system point spread function, including, but not limited to, the Richard-Lucy deconvolution algorithm and the Wiener filtering deconvolution algorithm.

[0018] The beneficial effects of this invention are: This invention utilizes a binarized amplitude-type spatial light modulator as the primary hardware encoding device. Compared to a DMD, this simplifies the optical system setup and allows for a more compact spatial structure. Therefore, the size of this invention is limited to within 300mm × 350mm × 200mm, making it easy to install on any optical platform. Furthermore, this invention can be directly combined and connected to various commercial microscope platforms, enabling these microscopes to achieve high-speed imaging exceeding camera frame rate limits.

[0019] This invention utilizes a three-step image reconstruction algorithm. The first step involves compressed sensing computation using a mask image and a measurement image. The second step involves a background removal algorithm to remove low-frequency background signals of biological fluorescence. The third step involves deconvolution to improve image resolution. Ultimately, this method can obtain high-speed, low-background, and high-resolution fluorescence microscopy images, suitable for various rapid in vivo imaging or high-throughput ex vivo imaging. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the optical system in one implementation of the present invention; Figure 2 This is a schematic diagram of the mechanical structure in one implementation of the present invention; Figure 3 This is a schematic diagram of the camera exposure timing and spatial light modulator mask loading timing in one implementation of the present invention; Figure 4 This is a schematic diagram of the system workflow in one implementation of the present invention; Figure 5 This is a schematic diagram of the three-step algorithm processing flow in one implementation of the present invention; Figure 6 This is a comparison diagram of the actual measured image and the results after each step of the algorithm processing in one implementation of the present invention; Figure 7 This is a comparison diagram of the actual measured image and the result after processing by the three-step algorithm in one implementation of the present invention; In the figure: 1. Commercial microscope platform; 2. Polarizing beam splitter; 3. Binarized amplitude spatial light modulator; 4. First relay lens; 5. First reflector; 6. Second relay lens; 7. Second reflector; 8. Camera. Detailed Implementation

[0021] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] The embodiments of the present invention and their specific processes are as follows: like Figure 1 As shown, the fluorescence of the commercial microscope platform 1 first propagates to the encoding modulation module, that is, to the polarization beam splitter 2, and is split into S-light and P-light of equal intensity at the polarization beam splitter 2. The S-light is reflected by the polarization beam splitter 2 and eventually leaves the system, while the P-light continues to propagate through the polarization beam splitter 2 and is imaged on the binary amplitude spatial light modulator 3. The binary amplitude spatial light modulator 3 loads different mask images in chronological order to encode the fluorescence image and converts the P-light into S-light, while reflecting it back to the polarization beam splitter 2 along the original path. The polarization beam splitter 2 reflects the S-light to the relay imaging module. like Figure 1 As shown, after the S-beam enters the relay imaging module, it is first deflected by the first relay lens 4. Then, the S-beam is reflected by the first reflector 5 to the second relay lens 6. After being deflected by the second relay lens 6, the S-beam is reflected by the second reflector 7 to the camera 8, where it is imaged. The binarized amplitude spatial light modulator 3 coincides with the front focal plane of the first relay lens 4, the rear focal plane of the first relay lens 4 coincides with the front focal plane of the second relay lens 6, and the rear focal plane of the second relay lens 6 coincides with the target surface of the camera 8. That is, the first relay lens 4 and the second relay lens 6 form a 4F system, and the binarized amplitude spatial light modulator 3 and the target surface of the camera 8 form a conjugate surface. Therefore, the S-beam is imaged exactly on the target surface of the camera 8.

[0023] like Figure 3In the imaging process shown, during a single exposure of camera 8, the binarized amplitude spatial light modulator 3 changes the mask multiple times, and the moment camera 8 begins exposure is exactly when the binarized amplitude spatial light modulator 3 begins loading the first mask. There are three ways to synchronize the timing of camera 8 and the binarized amplitude spatial light modulator 3: 1. The binarized amplitude spatial light modulator 3 sends a signal each time it begins loading the first mask, and camera 8 starts exposure upon receiving the signal; 2. Camera 8 sends a signal when it begins exposure, and the binarized amplitude spatial light modulator 3 starts loading the mask upon receiving the signal; 3. An external signal source is used to send a signal, and camera 8 and the binarized amplitude spatial light modulator 3 start exposure and mask loading respectively upon receiving the signal. In all three synchronization methods, the exposure time of camera 8 needs to be pre-set to be exactly equal to the time it takes for the binarized amplitude spatial light modulator 3 to load multiple masks.

[0024] like Figure 2 As shown, in one implementation of the present invention, the main body of the present invention is as follows: Figure 2 As shown on the right, the system includes a polarizing beam splitter 2, a binarized amplitude spatial light modulator 3, a first relay lens 4, a first reflecting mirror 5, a second relay lens 6, a second reflecting mirror 7, and a camera 8. All components are mounted on an optical support rod and fixed to an optical breadboard using pressure plates, and sealed with a correspondingly sized housing. The polarizing beam splitter 2 has a zoom sleeve installed near the commercial microscope platform 1 for connecting to various commercial microscope platforms and adjusting the focal point so that the fluorescence is imaged precisely on the target surface of the camera 8. Figure 2 The diagram on the left shows the connection between the main body of the present invention and the commercial microscope platform 1 in one implementation of the invention. The present invention uses a binary amplitude-type spatial light modulator 3 for fluorescence encoding. Its 180-degree reflectivity ensures that the incident and outgoing light paths coincide, resulting in a shorter optical path length and a more compact structure. The relay imaging module of the present invention uses two mirrors to further fold and compress the optical path, further reducing the overall size of the device. The main body's dimensions are only less than 300mm × 350mm × 200mm, resulting in a small space occupation and high integration.

[0025] like Figure 4 As shown, in one implementation of the present invention, the system workflow includes laser excitation of sample fluorescence, fluorescence imaging to a spatial light modulator, where the fluorescence is synchronously modulated by a mask, then acquired by a camera, and subsequently, an algorithm is used to reconstruct multiple background-free, high-resolution biofluorescence images from a single acquired image. Let biofluorescence be... ,in The horizontal and vertical coordinates on the image represent the point spread function of commercial microscope platforms that blurs the image of biological fluorescence when it is imaged onto a spatial light modulator. ,in Representing the convolution operator, mask modulation on a spatial light modulator can be expressed as the product of the mask and fluorescence, i.e. During a single camera capture, the mask changes multiple times, so the image captured by the camera can be represented as:

[0026] Considering the background fluorescence that may exist during the imaging process, the above equation needs to be rewritten as:

[0027] Where n represents the mask number, This indicates the total number of masks loaded during a single camera acquisition process. This represents background fluorescence. For algorithmic reconstruction, this refers to the fluorescence acquired... Reconstruction .

[0028] like Figure 5 As shown, the algorithm reconstruction part includes a compressed sensing solution unit, a background removal unit, and a deconvolution unit. For the compressed sensing solution unit, the above formula can be expressed as:

[0029]

[0030]

[0031]

[0032] The corresponding inverse problem is:

[0033] in, The prior constraints on the image can be represented by methods including, but not limited to, total variational priors, L1 norm priors, Hessian continuity priors, and weighted nuclear norm minimization priors. To solve the inverse problem, various inverse problem solving frameworks can be used, including but not limited to two-step iterative convergence algorithms, alternating direction multiplier methods, and generalized alternating projection methods. Simultaneously, solving this inverse problem requires N corresponding masks as input. The K images acquired by the camera are processed by a compressed sensing unit to generate K×N low-resolution fluorescence images with a background.

[0034] Next, the K×N images are processed by a background removal unit to generate K×N low-resolution fluorescence images without background. The background removal unit can use background removal algorithms including, but not limited to, wavelet transform-based and frequency domain filtering-based algorithms. Finally, a deconvolution unit is used to generate K×N high-resolution fluorescence images without background. The deconvolution unit can use, but not limited to, Richard Lucy deconvolution algorithms and Wiener filtering deconvolution algorithms. Simultaneously, to perform deconvolution calculations, the point spread function image of the system needs to be input, i.e. PSF image.

[0035] like Figure 6 The image shown is a three-dimensional fluorescence image of a mouse lung actually acquired and reconstructed in one implementation of the present invention. The xy and yz axis cross-sections of each image illustrate the role of each unit in the algorithm reconstruction step of the present invention. Specifically, the compressed sensing reconstruction unit enhances the image details along the z-axis, the background removal unit eliminates the influence of background fluorescence on image quality, resulting in higher image contrast, and the deconvolution unit improves image resolution, making its details clearer.

[0036] like Figure 7 The image shown is a comparison of the actual measured image and the result after processing by the three-step algorithm in one implementation of this method. In this experiment, one camera exposure corresponds to eight spatial light modulator mask transformations, so it is possible to reconstruct eight high-resolution images without background from one image. The comparison of details between the measured image and the reconstructed image shows that the algorithm reconstructs and restores details that were not present in the acquired image, especially at the arrow-marked area. The change of the lung tissue from never being connected to the connection can be clearly observed in the final reconstructed image, while this change cannot be distinguished in the acquired image. Figure 7 Part (a) represents an actual measured image in one implementation of this method. Figure 7 Part (b) indicates according to Figure 7 The result of processing the measurement image in part (a) using a three-step algorithm is eight high-resolution images without background that are ultimately reconstructed from a single measurement image. Figure 7 Part (c) is Figure 7 Enlarged view of some details in section (a) Figure 7 The middle (d) section is a magnified display of the corresponding reconstruction results.

[0037] In summary, this invention discloses an integrated compressed sensing high-speed imaging device and method. This invention can be directly integrated into a commercial microscope platform. Due to its rational and compact optical and mechanical structure design, this invention has a volume of only 300mm × 350mm × 200mm, resulting in low space occupancy. This invention utilizes optical hardware to encode and image fluorescence, and subsequently employs a three-step algorithm for image reconstruction, providing high-speed, low-background, and high-resolution imaging results for fluorescence microscopy.

Claims

1. An integrated compressed sensing high-speed imaging device, characterized in that, include: The coding and modulation module is used to encode and modulate biological fluorescence and transmit it to the next module; A relay imaging module is used to re-image the modulated biofluorescence onto the target surface of the camera (8). The decoding and reconstruction module is used to convert a single image measured by the camera into multiple high-resolution images.

2. The integrated compressed sensing high-speed imaging device according to claim 1, characterized in that: The encoding and modulation module includes a polarizing beam splitter (2) and a binary amplitude-type spatial light modulator (3). The direction of fluorescence propagation from the commercial microscope platform (1) is taken as the emission direction. The polarizing beam splitter (2) and the binary amplitude-type spatial light modulator (3) are arranged alternately in sequence. The fluorescence of the commercial microscope platform (1) first propagates to the polarizing beam splitter (2) and is divided into P light and S light of equal intensity. The P light continues to propagate through the polarizing beam splitter (2) to the binary amplitude-type spatial light modulator (3). This part of the fluorescence is encoded and modulated by the binary amplitude-type spatial light modulator (3), converted into S light and reflected back to the polarizing beam splitter (2). Then the polarizing beam splitter (2) reflects this part of the fluorescence to the relay imaging module.

3. The integrated compressed sensing high-speed imaging device according to claim 1, characterized in that: The relay imaging module includes a first relay lens (4), a first reflector (5), a second relay lens (6), a second reflector (7), and a camera (8). The first relay lens (4) is perpendicular to the direction of fluorescence emitted from the encoding imaging module. The first reflector (5) is arranged after the first relay lens (4) and rotates the direction of fluorescence counterclockwise by 90 degrees. The second relay lens (6) is arranged perpendicular to the direction of fluorescence propagation after the first reflector (5). The second reflector (7) is arranged after the second relay lens (6) and rotates the direction of fluorescence counterclockwise by 90 degrees again. The camera (8) is arranged perpendicular to the direction of fluorescence propagation after the second reflector (7). The fluorescence emitted from the encoding modulation module is first deflected by the first relay lens (4), then rotated 90 degrees counterclockwise by the first reflector (5), then deflected by the second relay lens (6), and then rotated 90 degrees counterclockwise by the second reflector (7), and finally imaged on the target surface of the camera (8).

4. The integrated compressed sensing high-speed imaging device according to claim 1, characterized in that: The encoding modulation module uses a binary amplitude spatial light modulator (3) to encode and modulate fluorescence. Its 180-degree reflectivity makes the incident light path and the outgoing light path coincide, so that the device has a shorter optical path length and a more compact structure. The relay imaging module uses two mirrors to further fold and compress the optical path, reducing the overall size of the device. The device is compatible with the commercial microscope platform (1) and receives fluorescence collected from the commercial microscope platform (1). It does not need to be equipped with a microscope objective or other fluorescence collection devices.

5. The integrated compressed sensing high-speed imaging device according to claim 1, characterized in that: The decoding and reconstruction module includes a compressed sensing solution unit, a background removal unit, and a deconvolution unit; The input to the compressed sensing solution unit is an coded mask pattern loaded on a binary amplitude spatial light modulator (3) and a measurement image from a camera (8), and the output is multiple low-resolution fluorescence images. The input to the background removal unit is multiple low-resolution fluorescence images output by the compressed sensing solution unit, and the output is multiple low-resolution fluorescence images without background. The input to the deconvolution unit is multiple low-resolution fluorescence images without background and a system point spread function image output by the background removal unit, and the output is multiple high-resolution fluorescence images without background. The system point spread function is a response function describing the optical imaging system to imaging of an ideal point light source. The system point spread function image is obtained by imaging the fluorescent microspheres during the system calibration stage. A measurement image obtained by the camera (8) is sequentially processed by the compressed sensing solution unit, the background removal unit and the deconvolution unit to obtain multiple high-resolution fluorescence images without background.

6. An integrated compressed sensing high-speed imaging method, characterized in that, This method is used to implement the apparatus of any one of claims 1-5, specifically comprising: Bioluminescence emitted from the commercial microscope platform (1) first passes through a polarizing beam splitter (2). P-light propagates through the polarizing beam splitter (2) to a binary amplitude-modulated spatial light modulator (3). The binary amplitude-modulated spatial light modulator (3) encodes and modulates the fluorescence, converting the P-light into S-light. The S-light is then reflected back to the polarizing beam splitter (2) along the same path. The polarizing beam splitter (2) reflects the S-light to a first relay lens (4). The fluorescence passing through the first relay lens (4) is reflected by a first reflecting mirror (5) to a second relay lens (6). The fluorescence of the second relay lens (6) is reflected by the second mirror (7) onto the camera (8) to complete the imaging process. The measurement image obtained on the camera (8) and the coded mask pattern on the binarized amplitude spatial light modulator (3) are input into the compressed sensing solution unit to obtain multiple low-resolution fluorescence images with background. The multiple low-resolution fluorescence images with background are input into the background removal unit to obtain multiple low-resolution fluorescence images without background. The multiple low-resolution fluorescence images without background and the system point spread function image are input into the deconvolution unit to obtain multiple high-resolution fluorescence images without background.

7. The integrated compressed sensing high-speed imaging method according to claim 6, characterized in that: The aforementioned binary amplitude spatial light modulator (3) can load any binary image; At the position where the pixel value is 1, the P light is converted into the S light, which is reflected to the polarizing beam splitter (2) and then reflected to the first relay lens (4). At the position where the pixel value is 0, the P light is not converted into the S light, and after being reflected, it passes through the polarizing beam splitter (2) and finally leaves the system.

8. The integrated compressed sensing high-speed imaging method according to claim 6, characterized in that: The binary amplitude spatial light modulator (3) is arranged at the biofluorescence image plane of the commercial microscope platform (1); the first relay lens (4) and the second relay lens (6) form a 4F system, and its front focal point coincides with the binary amplitude spatial light modulator (3); the camera (8) is arranged at the back focal point of the 4F system, so the biofluorescence is exactly imaged on the camera (8), and the binary amplitude spatial light modulator (3) and the camera (8) form a conjugate plane.

9. The integrated compressed sensing high-speed imaging method according to claim 6, characterized in that: The binary amplitude spatial light modulator (3) needs to be precisely synchronized with the camera (8). Depending on the synchronization method, the two can receive synchronization signals from a unified signal source, or the binary amplitude spatial light modulator (3) can send synchronization signals and the camera (8) can receive the synchronization signals or the camera (8) can send synchronization signals and the binary amplitude spatial light modulator (3) can receive the synchronization signals. When the camera (8) receives the signal and starts exposure or starts exposure autonomously, the binary amplitude spatial light modulator (3) starts loading the mask. According to the pre-set control program, the binary amplitude spatial light modulator (3) changes the mask multiple times during one exposure of the camera (8).

10. The integrated compressed sensing high-speed imaging method according to claim 6, characterized in that: The compressed sensing solution unit is composed of various inverse problem solution frameworks and various image priors. The inverse problem solution frameworks include, but are not limited to, two-step iterative convergence algorithms, alternating direction multiplier methods, and generalized alternating projection methods. The various image priors include, but are not limited to, total variational priors, L1 norm priors, Hessian continuity priors, and weighted kernel norm minimization priors. The background removal unit consists of various image background removal algorithms, including, but not limited to, background removal algorithms based on wavelet transform and background removal algorithms based on frequency domain filtering. The deconvolution unit consists of various algorithms that improve the resolution of the original image by using the system point spread function, including, but not limited to, the Richard-Lucy deconvolution algorithm and the Wiener filtering deconvolution algorithm.

Images