A system for realizing double-channel self-interference digital holographic real-time edge enhancement by T-Net

Abstract

The application relates to the technical field of optical imaging, and particularly discloses a system for realizing real-time edge enhancement of double-channel self-interference digital holography through T-Net, which comprises the following steps: a double-light-source unit irradiates a target object through a collimating lens; a shared imaging lens collects reflected light; a spatial light modulator is loaded with a vortex phase for edge modulation; a half-wave plate and a polarization beam splitter are used for beam splitting to form orthogonal polarization double channels; double CMOS image sensors synchronously collect holograms with a phase shift difference of pi; a data processing unit reconstructs an edge-enhanced image by eliminating a background term and a conjugate image and an angular spectrum propagation function through T-Net neural network parameters pre-stored in a memory and combined with a three-step phase shift algorithm to complete a third pi / 2 phase shift graph.

Description

Technical Field

[0001] This invention belongs to the field of optical imaging technology, and in particular relates to a system for real-time edge enhancement of dual-channel self-interference digital holography using T-Net. Background Technology

[0002] Edge information is crucial in image processing. It not only carries the main carriers of object shape, structure, and spatial features but also directly affects the accuracy of target recognition and segmentation. Incoherent Digital Holography (IDH) edge enhancement is a holographic imaging technique based on incoherent light sources. Unlike traditional holography, which requires interference between object and reference beams, IDH is an incoherent 3D optical imaging technique that can directly record holograms from sensors without scanning or mechanical movement. During numerical reconstruction, different depths of object layers can be recovered using diffraction propagation algorithms to achieve edge enhancement in 3D imaging. However, current implementations often require multiple exposures, i.e., recording multiple phase-shifted holograms to remove zero-order and conjugate images. This method increases imaging time and operational complexity. Therefore, designing a high-quality IDH edge enhancement system that can be performed with a single exposure is highly desirable.

[0003] Current edge enhancement techniques in incoherent digital holography primarily involve introducing high-order vortex phase modulation into the system to achieve flexible control of edge width and enhance edge feature recognition. Improved vortex angular spectrum algorithms are used in IDH systems, simultaneously adjusting amplitude and phase during numerical reconstruction, improving edge sharpness and providing direction selectivity and aberration adaptation. Bessel-like spiral phase modulation is also applied to IDH systems, effectively suppressing sidelobe interference from vortex light fields and significantly improving background suppression and edge contrast in 3D reconstructed images. Simultaneously, computational methods based on synthetic point spread functions (PSFs) have been proposed, using iterative optimization to reconstruct PSFs with edge enhancement properties, enabling edge extraction without relying on complex optical modulation. These methods exhibit excellent edge enhancement performance under incoherent imaging conditions, expanding their application potential in fields such as biological microscopy and target recognition. While these methods have made progress in image quality and controllability, current implementations often require multiple exposures, i.e., recording multiple phase-shifted holograms to remove zero-order and conjugate images. This increases imaging time and operational complexity. In addition, the accumulation of environmental noise during multiple exposures leads to a decrease in contrast, affecting the final image quality.

[0004] Therefore, achieving a single-exposure system that simultaneously achieves high temporal resolution and high image quality remains a challenge for this technology. Summary of the Invention

[0005] The purpose of this invention is to provide a system for real-time edge enhancement of dual-channel self-interference digital holography using T-Net, so as to achieve the technical effects of high temporal resolution real-time imaging, elimination of conjugate images and zero-order terms, and improvement of edge contrast and reconstruction accuracy.

[0006] To solve the above-mentioned technical problems, the specific technical solution of the present invention is as follows:

[0007] In some embodiments of this application, a system for real-time edge enhancement of dual-channel self-interference digital holography using T-Net is provided, comprising:

[0008] A dual-light source unit, comprising a first light-emitting diode, a second light-emitting diode, a first collimating lens, and a second collimating lens, wherein the first collimating lens is located in the output light path of the first light-emitting diode, and the second collimating lens is located in the output light path of the second light-emitting diode;

[0009] The first target object is located in the collimated light illumination area of ​​the first collimating lens;

[0010] The second target object is located in the collimated light illumination area of ​​the second collimating lens;

[0011] A shared imaging lens, wherein the incident end of the shared imaging lens receives reflected light from a first target object and a second target object;

[0012] A spatial light modulator, wherein the spatial light modulator is located in the outgoing light path of a shared imaging lens;

[0013] A half-wave plate, wherein the half-wave plate is located on the output light path of spatial light modulation, and a polarizing beam splitter is provided on the output light path;

[0014] The polarization beam splitter is provided with a first output optical path and a second output optical path, and a first CMOS image sensor and a second CMOS image sensor are respectively provided on them;

[0015] The data input interface is connected to the output terminals of the first CMOS image sensor and the second CMOS image sensor, and is electrically connected to the memory and the processor.

[0016] In some embodiments of this application, the center wavelength range of the first light-emitting diode and the second light-emitting diode is 615nm to 635nm.

[0017] In some embodiments of this application, the angle between the fast axis direction of the half-wave plate and the y-axis is 22.5°±0.5°.

[0018] In some embodiments of this application, the T-Net neural network includes an encoder and a decoder, wherein: the encoder consists of 4 convolutional layers, each followed by a max pooling layer; and the decoder consists of 4 deconvolutional layers, each followed by an upsampling layer.

[0019] In some embodiments of this application, the pixel size of the first CMOS image sensor and the second CMOS image sensor is from 1.4μm×1.4μm to 3.5μm×3.5μm.

[0020] In some embodiments of this application, the focal length of the common imaging lens (L3) is 75mm to 150mm.

[0021] In some embodiments of this application, the fast axis direction of the half-wave plate forms an angle of 17.5° to 27.5° with the y-axis.

[0022] Compared with existing technologies, the advantages of this invention are as follows: By employing a dual-channel synchronous acquisition unit and a single-exposure design, it simultaneously acquires two holograms with a phase shift difference of π, overcoming the imaging delay problem caused by traditional multiple exposures. The temporal resolution is improved to the millisecond level, meeting the real-time edge enhancement requirements of dynamic targets. Combined with the directional edge enhancement characteristics of the vortex phase modulation unit and the third phase-shifted image completed by the T-Net neural network, the zero-order term and conjugate image can be completely suppressed through a three-step phase-shifting algorithm. The parameter tolerance design of the half-wave plate angle range and CMOS pixel size reduces the difficulty of optical assembly and adjustment; the lightweight T-Net structure is compatible with conventional processors, eliminating the need for dedicated computing hardware. By adjusting the angular spectrum propagation distance, the system can achieve layered edge enhancement reconstruction for targets at different axial depths, making it suitable for three-dimensional scenes such as biological microscopy. Attached Figure Description

[0023] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0024] Figure 1 This is a schematic diagram of the principle of dual-channel incoherent digital holography provided in an embodiment of the present invention;

[0025] Figure 2 This is a schematic diagram of the system flow for implementing real-time edge enhancement of dual-channel incoherent digital holography using T-Net, as provided in an embodiment of the present invention.

[0026] Figure 3 This is a schematic diagram of the experimental device for a dual-channel incoherent digital holographic system provided in an embodiment of the present invention;

[0027] Figure 4This is a schematic diagram of a set of results from the verification data provided in an embodiment of the present invention;

[0028] Figure 5 (a) Schematic diagram of the edge enhancement reconstruction results provided in the embodiments of the present invention;

[0029] Figure 6 A schematic diagram illustrating the comparison of partial reconstruction results of the validation set provided in an embodiment of the present invention;

[0030] Figure 7 This is a schematic diagram of the edge imaging results of two targets at different axial positions provided in an embodiment of the present invention.

[0031] in, Figure 1 The figures below are (a) the optical device, and (b) the phase-shifted holograms and Jones matrix of the two channels, respectively.

[0032] Figure 4 (a) shows the phase-shifted interferogram and network output, (b) shows the phase-shifted interferogram, and (c) shows the cross-sectional intensity curve of column 336 in the interferograms of (a) and (b). The scale bar in the figure is 300 μm.

[0033] Figure 5 (a) shows the results of the two-step phase-shift reconstruction, (b) shows the results of the reconstruction using a 1-to-2 U-Net, and (c) shows the results using the method presented in this paper. The scale bar in the figures is 300 μm.

[0034] Figure 6 (a) shows the two-step phase-shift reconstruction result, (b) shows the reconstruction result using 1-to-2 U-Net, and (c) shows the result using the method presented in this paper. The scale bar in the figure is 300 μm.

[0035] Figure 7 In the figures (a) and (b), targets 1 and 2 are reconstructed at different axial positions. The reconstruction distance is focused at the white arrow object. The scale bar in the figure is 300 μm. Detailed Implementation

[0036] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0037] To better understand the purpose, structure, and function of this invention, the invention will be described in further detail below with reference to the accompanying drawings.

[0038] Example 1

[0039] This application discloses a real-time high-precision edge enhancement method for dual-channel incoherent digital holography based on T-Net, comprising the following steps:

[0040] System Setup: A dual-channel self-interference digital holographic system is constructed using two identical light sources (LEDs with a center wavelength of 625nm) to illuminate two target objects. The light is collimated by front lenses L1 and L2 before illuminating the target. The light reflected from the target is collected by a shared imaging lens L3 and then phase-modulated by a spatial light modulator (SLM) with a vortex phase. The dual-channel synchronous phase shift is achieved using a half-wave plate (HWP) and a polarizing beam splitter (PBS). The two orthogonally polarized modulated beams are synchronously acquired by two identical CMOS cameras to generate two holograms with phase shift differences, enabling the acquisition of dual-channel phase-shifted images in a single exposure.

[0041] Data Collection: Based on the aforementioned optical path system, this invention collected 3000 sets of holographic data. Each set contains three phase-shifted interferograms, specifically three frames with phase shift angles of 0, π / 2, and π. The images with phase shifts of 0 and π were directly obtained from the experimental system, while the image with a phase shift of π / 2 was used as the training target for the T-Net network. The dataset was divided into training, validation, and test sets in an 8:1:1 ratio to ensure the stability of network training and the model's generalization ability.

[0042] Network Construction and Training: To complete the third phase-shifted image (π / 2), a deep neural network with a T-Net structure was used. The network architecture referenced U-Net, consisting of a symmetrical encoder and decoder, and includes multiple layers of convolution, pooling, and upsampling operations for extracting and reconstructing holographic features. The network input consists of two holograms with phase shifts of 0 and π, and the output is the predicted π / 2 phase-shifted image. The root mean square error (RMSE) loss function was chosen to ensure the accuracy of the predicted image. After training, the network can be used as a virtual phase shifter in practical applications to generate a complete three-phase-shifted image in a single exposure.

[0043] Testing the network model: The trained T-Net model was applied to the test set. A third predicted image (π / 2 phase shift) was generated by inputting two holograms with phase shifts of 0 and π. This predicted result was compared with the π / 2 phase shift image acquired in actual experiments to verify the accuracy and consistency of the phase shift image output by the network in terms of intensity distribution and structure reconstruction, thereby evaluating the network's reconstruction capability and generalization performance.

[0044] Verification of reconstruction results: To further verify the usability of the phase shift map generated by the network in actual imaging tasks, the π / 2 phase shift map output by the network was compared with the acquired 0 and π phase shifts. Figure 1In the same input three-step phase-shift reconstruction algorithm, the complex amplitude hologram of the object is recovered. The reconstructed image is imaged using the angular spectrum propagation algorithm to obtain the edge-enhanced image result. By comparing the results with those of two-step phase-shift reconstruction and three-step phase-shift reconstruction using a 1-to-2U-Net network, structural similarity (SSIM), peak signal-to-noise ratio (PSNR), and other indicators are calculated to quantitatively evaluate the consistency between the reconstruction quality of the network output image and the real image, demonstrating that the present invention has high accuracy and real-time performance in edge enhancement reconstruction.

[0045] The technical effects achieved by the above technical solution in the embodiments of this application are as follows:

[0046] By constructing a dual-channel incoherent digital holographic system, two vortex holograms with different phase shifts can be simultaneously acquired in a single exposure. Secondly, a T-Net is built to predict a third phase-shifted vortex hologram, and then a three-step phase-shifting algorithm and a backpropagation algorithm are combined to achieve image edge enhancement and reconstruction. This network only needs to handle the mapping relationship between holograms with different phase shifts, exhibiting higher accuracy and better generalization ability. The trained network can act as a phase shifter, generating holograms with specific phase shifts from holograms recorded in a single exposure, which greatly saves the acquisition time of phase-shifted holograms. Network analysis and optical experimental results verify the effectiveness of the proposed method, providing a new scheme for fast incoherent digital holographic edge enhancement imaging.

[0047] Example 2

[0048] This application adopts some of the technical features from the above embodiments, wherein, see appendix. Figure 1-7 As shown,

[0049] Figure 1 A schematic diagram of a dual-channel incoherent digital holographic system is shown, which combines a dual-channel synchronous phase-shift setup with a self-interference digital holographic system. A self-emitting point source is located at z0 on a lens L (focal length f). L is used to collect and collimate the light from the object point. For ease of analysis, the complex amplitude of the diffracted light field at (x0, y0, z0) is explored using scalar diffraction theory. The distance between the SLM and L is d, and the complex amplitude of the light field on the front plane of the SLM can be expressed as:

[0050]

[0051] In the formula Let be the horizontal coordinate of the target plane. For the plane coordinates of the SLM, The value is a complex constant related to the location of the point source, and "*" indicates a two-dimensional spatial convolution operation.

[0052] in the formula and Q s(1 / z)=exp(iπ(x s 2 +y s 2 ) / λz), representing the linear phase function and the quadratic phase function, respectively.

[0053] The SLM's working axis is aligned with the horizontal direction (i.e., the y-axis), and the polarizer P is oriented at an angle α relative to the y-axis. Under these conditions, the polarization component of the light emitted from the object along the y-axis will be subject to a quadratic phase factor applied to the SLM. The modulation is performed, at which point a phase mask with a focal length of f is applied to the SLM. a The polarization component along the vertical direction (i.e., the x-axis) is not modulated. The light field in the plane after passing through an SLM (with a phase mask loaded on the SLM at focal length f0) This can be viewed as a vector superposition of linearly polarized spherical waves with different curvatures along the x-axis and y-axis, and can be described using the Jones matrix as follows:

[0054]

[0055] The dual-channel synchronous phase-shifting device mainly consists of a half-wave plate (HWP), a polarizing beam splitter (PBS), and two identical CMOS cameras. The fast axis of the HWP is tilted at 22.5° relative to the y-axis. After passing through the polarizing beam splitter, the polarization directions of the two channels are 0° and 90° relative to the y-axis, respectively. Based on the Jones matrix of typical optical elements, the polarization modulation process of the dual-channel synchronous phase-shifting device is described using the Jones matrix form. Figure 1 (b) shows the phase-shifted holograms of the two channels and the Jones matrix of the polarization elements. In the analysis of the optical field, only the polarization state changes and phase delays caused by the optical components are considered, while the spatial phase changes caused by diffraction propagation are ignored. Therefore, the expression does not contain coordinate dependencies of the optical field expression. Thus, the Jones matrix expressions for the optical fields on camera 1 and camera 2 are obtained as follows:

[0056]

[0057] By configuring the system, a pair of holograms with a π phase shift can be acquired simultaneously from two cameras. CAM1 and I CAM2 Their intensity expressions are:

[0058]

[0059] Based on the above analysis, object light The light is split into a pair of orthogonally linearly polarized beams by the SLM. When the phase mask loaded onto the SLM is a vortex lens phase, the modulation expression is:

[0060]

[0061] The vortex phase distribution is as follows (l is the angular momentum number;) (where the angle is in polar coordinates on the Fourier plane), then the two beams of light propagate a distance z. h The light reaches the camera plane through the polarizer. At this point, the projections of two beams of light with different curvatures along the polarization axis will interfere with each other, forming a hologram. Combining equation (2), the intensity distribution of the recorded point source hologram can be expressed as:

[0062]

[0063] Where θ is the phase shift introduced by the dual-channel synchronous phase shift setting, Q(1 / z) h )=exp(iπ(x s 2 +y s 2 ) / λz h ) represents a quadratic phase function. and These are the lateral coordinates of the camera plane and the image plane, respectively. Due to insufficient acquired information, the two-step phase shift algorithm has weak suppression capabilities against random noise, easily leaving background terms and conjugate images, leading to a decrease in the contrast of the reconstructed image. Therefore, a T-Net is constructed to generate a third phase-shifted hologram, and a three-step phase shift algorithm is combined to eliminate background terms and conjugate images.

[0064] The technical effects achieved by the above technical solution in the embodiments of this application are as follows:

[0065] By constructing a dual-channel incoherent digital holographic imaging system, two vortex holograms with different phase shifts can be acquired simultaneously during a single exposure. Based on the two acquired holograms, a third vortex hologram with the target phase shift is predicted using a T-Net neural network. After acquiring three holograms with different phases, a three-step phase-shifting algorithm and a backpropagation algorithm are further introduced to achieve edge enhancement and reconstruction of the holographic image. The T-Net network only needs to learn the mapping relationship between holograms with different phase shifts, thus exhibiting high reconstruction accuracy and good generalization ability. After network training, the network model can be used as a virtual phase shifter, generating holograms with specific phase shifts from holograms acquired in a single exposure without the need for mechanical phase-shifting components, thereby significantly reducing the time cost required for hologram acquisition. Through network performance evaluation and optical experimental verification, the method proposed in this invention demonstrates excellent performance in image edge enhancement, reconstruction quality, and processing efficiency, providing an efficient and feasible new technical solution for fast incoherent digital holographic edge enhancement imaging.

[0066] In the description of this application, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0067] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0068] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0069] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0070] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A system for real-time edge enhancement of dual-channel self-interference digital holography using T-Net, characterized in that, include: A dual-light source unit, comprising a first light-emitting diode, a second light-emitting diode, a first collimating lens, and a second collimating lens, wherein the first collimating lens is located in the output light path of the first light-emitting diode, and the second collimating lens is located in the output light path of the second light-emitting diode; The first target object is located in the collimated light illumination area of ​​the first collimating lens; The second target object is located in the collimated light illumination area of ​​the second collimating lens; A shared imaging lens, wherein the incident end of the shared imaging lens receives reflected light from a first target object and a second target object; A spatial light modulator, wherein the spatial light modulator is located in the outgoing light path of a shared imaging lens; A half-wave plate, wherein the half-wave plate is located on the output light path of spatial light modulation, and a polarizing beam splitter is provided on the output light path; The polarization beam splitter is provided with a first output optical path and a second output optical path, and a first CMOS image sensor and a second CMOS image sensor are respectively provided on them; A data input interface is provided, which is connected to the output terminals of the first CMOS image sensor and the second CMOS image sensor, and is electrically connected to the memory and the processor. The T-Net neural network includes an encoder and a decoder, wherein the encoder consists of four convolutional layers, each followed by a max pooling layer; The decoder consists of four deconvolutional layers, each followed by an upsampling layer; Multiple sets of phase-shift interferograms were acquired using a dual-channel system. Each set included three images with phase shifts of 0, π / 2, and π. The 0 and π phase-shift images were directly acquired by the system, while the π / 2 phase-shift image was used as the training target for the T-Net network. The dataset was divided into training, validation, and test sets proportionally. A T-Net model was trained using the dataset, with the 0 and π phase-shift images as input and the predicted π / 2 phase-shift image as output. The root mean square error (RMSE) loss function was used. The 0 and π phase-shift images acquired by the dual channels were input into the trained T-Net model to predict and generate the π / 2 phase-shift image. The predicted result was compared with the π / 2 phase-shift image acquired in the actual experiment to verify the accuracy and consistency of the phase-shift image output by the T-Net network in terms of intensity distribution and structure reconstruction. The π / 2 phase-shift image output by the T-Net network and the acquired 0 and π phase-shift images were input together into a three-step phase-shift reconstruction algorithm to recover the complex amplitude hologram of the object. The reconstructed image was then imaged using an angular spectrum propagation algorithm to obtain the edge-enhanced image result.

2. The system for real-time edge enhancement of dual-channel self-interference digital holography using T-Net as described in claim 1, characterized in that, The first and second light-emitting diodes have a center wavelength range of 615nm to 635nm.

3. The system for real-time edge enhancement of dual-channel self-interference digital holography using T-Net as described in claim 1, characterized in that, The fast axis of the half-wave plate makes an angle of 22.5° ± 0.5° with the y-axis.

4. The system for real-time edge enhancement of dual-channel self-interference digital holography using T-Net as described in claim 1, characterized in that, The pixel size of the first CMOS image sensor and the second CMOS image sensor is from 1.4μm×1.4μm to 2.4μm×2.4μm.

5. A system for real-time edge enhancement of dual-channel self-interference digital holography using T-Net as described in claim 1, characterized in that, The focal length of the shared imaging lens (L3) is 75mm to 150mm.

6. A system for real-time edge enhancement of dual-channel self-interference digital holography using T-Net as described in claim 1, characterized in that, The fast axis of the half-wave plate forms an angle of 17.5° to 27.5° with the y-axis.

Images