• Photonics Research
  • Vol. 9, Issue 3, 03000B57 (2021)
Kangning Zhang, Junjie Hu, and Weijian Yang*
Author Affiliations
  • Department of Electrical and Computer Engineering, University of California, Davis, California 95616, USA
  • show less

    Abstract

    The need for high-speed imaging in applications such as biomedicine, surveillance, and consumer electronics has called for new developments of imaging systems. While the industrial effort continuously pushes the advance of silicon focal plane array image sensors, imaging through a single-pixel detector has gained significant interest thanks to the development of computational algorithms. Here, we present a new imaging modality, deep compressed imaging via optimized-pattern scanning, which can significantly increase the acquisition speed for a single-detector-based imaging system. We project and scan an illumination pattern across the object and collect the sampling signal with a single-pixel detector. We develop an innovative end-to-end optimized auto-encoder, using a deep neural network and compressed sensing algorithm, to optimize the illumination pattern, which allows us to reconstruct faithfully the image from a small number of measurements, with a high frame rate. Compared with the conventional switching-mask-based single-pixel camera and point-scanning imaging systems, our method achieves a much higher imaging speed, while retaining a similar imaging quality. We experimentally validated this imaging modality in the settings of both continuous-wave illumination and pulsed light illumination and showed high-quality image reconstructions with a high compressed sampling rate. This new compressed sensing modality could be widely applied in different imaging systems, enabling new applications that require high imaging speeds.

    1. INTRODUCTION

    Landscape of imaging methods using a single-pixel detector. (a) Point scanning system where the signal from an individual pixel is sequentially recorded. (b) A conventional single-pixel camera where different patterns are sequentially projected on the entire object, and the overlap integrals between the object and each pattern are measured. (c) Deep compressed imaging via optimized pattern scanning (DeCIOPS), where a pattern is scanned across the object, and the subsampled convolution between the pattern and the object is measured.

    Figure 1.Landscape of imaging methods using a single-pixel detector. (a) Point scanning system where the signal from an individual pixel is sequentially recorded. (b) A conventional single-pixel camera where different patterns are sequentially projected on the entire object, and the overlap integrals between the object and each pattern are measured. (c) Deep compressed imaging via optimized pattern scanning (DeCIOPS), where a pattern is scanned across the object, and the subsampled convolution between the pattern and the object is measured.

    In this paper, we propose and demonstrate a deep compressed sensing modality, which can significantly increase the imaging speed while preserving a high reconstruction quality. This approach combines the strength of both compressed sensing and point-scanning imaging, and we call it deep compressed imaging via optimized pattern scanning (DeCIOPS). Instead of projecting multiple binary patterns onto the entire object sequentially, we utilize only one gray-scale optimized pattern and project it to a small subset of the object. We then scan the pattern across the object by using fast scanning mirrors and collect the signal convolutionally using a single-pixel detector [Fig. 1(c)]. Compared with the conventional single-pixel camera, which relies on sequentially switching the sampling pattern on a DMD, our scanning approach significantly increases the sampling speed. Compared with the point-scanning system, our method samples a much larger portion of the object at once and recovers the resolution computationally. This allows a great reduction of the sampling number and thus increases the frame rate. We note that the improvement of imaging speed does not require an increase of light energy. In fact, the required light dosage in our method is smaller than in the conventional point-scanning system due to a reduced number of measurements. We build an auto-encoder framework [17] to optimize the sampling pattern. The image acquisition system is treated as an encoder, where the high-resolution object is encoded through the sampling pattern into a few measurements. We then formulate an iterative shrinkage-thresholding algorithm network (ISTA-Net) [18], a CS-induced neural network inspired by the iterative shrinkage-thresholding algorithm (ISTA) [19], as a decoder to reconstruct the image. This auto-encoder is trained in an end-to-end fashion. Such a framework can learn an optimized sampling pattern and simultaneously recover a high-resolution image by extracting the feature of sparsity and searching the optimal pair of the encoder and decoder with the lowest incoherence [11], which is one key feature of DeCIOPS versus other compressed sensing or deep-learning-based super-resolution imaging modalities [2023]. This new imaging modality can be accustomed to any light-scanning imaging system and will greatly benefit the high-throughput imaging applications.

    This paper is organized as follows. In Section 2, we introduce the mechanisms of the imaging modality in two configurations of illuminations, continuous-wave (CW) and pulsed light source, as well as the auto-encoder framework and the deep compressed sensing neural network for optimizing the imaging and reconstruction. In Sections 3 and 4, we show the simulation results and experimental results. In Section 5, we discuss the system performance under different signal-to-noise ratio (SNR) and compression ratio and how DeCIOPS can be applied in two-photon microscopy and passive lighting conditions.

    2. PRINCIPLE

    A. Image Formation

    Schematic of the undersampling schemes in DeCIOPS. (a) Conventional pixel-by-pixel point scanning. (b) Pixel-by-pixel point scanning with a simple undersampling scheme. (c) DeCIOPS in a CW light source configuration with an illumination pattern of a uniform mask (left) or an optimized mask (right). (d) DeCIOPS in a low-repetition-rate pulsed light source configuration with a uniform mask (left) or an optimized mask (right) as an illumination pattern. The mathematic formula below each panel illustrates the process of image formation, where g1 and g2 are both square shaped.

    Figure 2.Schematic of the undersampling schemes in DeCIOPS. (a) Conventional pixel-by-pixel point scanning. (b) Pixel-by-pixel point scanning with a simple undersampling scheme. (c) DeCIOPS in a CW light source configuration with an illumination pattern of a uniform mask (left) or an optimized mask (right). (d) DeCIOPS in a low-repetition-rate pulsed light source configuration with a uniform mask (left) or an optimized mask (right) as an illumination pattern. The mathematic formula below each panel illustrates the process of image formation, where g1 and g2 are both square shaped.

    Mathematically, in low-repetition-rate pulsed light illumination, the image formation of DeCIOPS can be expressed as b=fn×n(x*g),where x is the object, g is a square-shaped illumination pattern, * represents the 2D convolution, fn×n(·) models the n×n undersampling, and b is the measured image. Here, we assume that the mask has a size of n×n pixels. In the particular case where g is uniform mask g1 [Fig. 2(d), left], Eq. (1) is equivalent to a naïve undersampling by unweighted averaging of every n×n pixels of the full resolution image [Fig. 2(a)] acquired in single-point scanning. As discussed in Section 2.B, g can be optimized to achieve the best image reconstruction performance [Fig. 2(d), right].

    In the CW light or high-repetition-rate pulsed light illumination case, where the detector continuously integrates the signal, we configure the illumination pattern in a size of n×1 pixels. When this pattern sweeps n columns, the information of n×n pixels is integrated into a single measurement [Fig. 2(c)]. We can use the same mathematical formulation as Eq. (1) to model the image formation, where each column in the n×n mask g is identical.

    B. End-to-End Optimized Auto-Encoder Framework

    End-to-end optimized auto-encoder framework of image formation and reconstruction in DeCIOPS. The encoder models the image formation. It encodes the high-resolution (HR) object x into a low-resolution (LR) output b+ through subsampled convolution Φ and additive noise. The decoder is implemented with an ISTA-Net, which contains N phases and reconstructs the object x(N). Each phase is realized by a structure-symmetric pair of a forward transform F(k) and a backward transform F−1(k) with a soft shrinkage threshold, which factually matches one iteration in the conventional ISTA. ReLU, rectified linear unit; Soft(·), soft shrinkage threshold.

    Figure 3.End-to-end optimized auto-encoder framework of image formation and reconstruction in DeCIOPS. The encoder models the image formation. It encodes the high-resolution (HR) object x into a low-resolution (LR) output b+ through subsampled convolution Φ and additive noise. The decoder is implemented with an ISTA-Net, which contains N phases and reconstructs the object x(N). Each phase is realized by a structure-symmetric pair of a forward transform F(k) and a backward transform F1(k) with a soft shrinkage threshold, which factually matches one iteration in the conventional ISTA. ReLU, rectified linear unit; Soft(·), soft shrinkage threshold.

    The decoder takes b+ as the input and aims to reconstruct the original object x by solving the following convex optimization problem with a constraint of the sparse representation of x: argminx12Φxb+22+λΨ(x)1,where Ψ(x) denotes a transform of x into a sparse representation under the basis of Ψ, and λ is a hyperparameter.

    The solution of the problem in Eq. (3) can be initialized by calculating the pseudo-inverse of the encoder from the measurement b+. We then use ISTA [19] to find an optimized solution of x by iterating the following two steps: r(k)=x(k1)ρΦT(Φx(k1)b+),x(k)=argminx12xr(k)22+λΨ(x)1,where k denotes the kth ISTA iteration step, and ρ is the step size.

    As Ψ(·) is predefined empirically and may not be suitable for the data, we adopt the ISTA-Net algorithm [18], which can learn Ψ(·) through the data. In ISTA-Net, Ψ(·) is replaced with a trainable neural network F(·), and the optimization problem in Eq. (3) can be rephrased into the following L1-norm regularization problem with a nonlinear transform F: argminx12Φxb+22+λF(x)1.

    The kth iteration step in the original ISTA is replaced by a series of symmetric learnable parameters in the kth ISTA-Net phase: F(x(k))=soft(F(r(k)),θ),x(k)=F1(soft(F(r(k)),θ)),where θ is a learnable parameter in the kth module, F1(·) is the inverse of F(·), and soft(·) represents the soft shrinkage threshold. Finally, we obtain the output of the decoder x(N) after a total number of N ISTA-Net phases.

    The loss function of ISTA-Net is obtained by calculating the mean square error (MSE) between the output and the ground truth with the constraint of F1°F=I, as both F1 and F are learnable and symmetric, where I is the identity operator. As a result, we have the following loss function with the symmetry constraint: Ltotal=Lerror+γLconstraint=x(N)x22γ(k=1NF1(k)(F(k)(x))x22),where γ is the weight of the symmetry constraint.

    3. SIMULATION RESULTS

    We trained the auto-encoder using 1500 samples of natural scenes (2D gray-scale image, 256×256 pixel size) from ImageNet [24] and validated the model with 79 samples from two widely used benchmark datasets: Set11 [25] and BSD68 [26]. As an illustration, we chose an undersampling rate of 6.25% (4×4 undersampling) and initialized the pattern g as a 4×4 normalized random Gaussian matrix. γ was set to be 0.01 in the loss function, accompanied by Adam optimization with a learning rate of 1×104. We included additive noises in the measurement (5%–10% of the signal strength). The training was performed on a GPU RTX2080Ti 11 GB. The training work of N=9 ISTA-Net phases takes 5  h for 200 epochs with a batch size of five. We trained two independent auto-encoders, one with a constraint on g so each column of g is identical (CW light or high-repetition-rate pulsed light illumination) and one without such a constraint (low-repetition-rate pulsed light illumination). The reconstruction results are evaluated in terms of peak signal-to-noise ratio (PSNR) and spatial resolution by using Fourier ring correlation [27] in the validation dataset. As a control measure, we compared the reconstruction performance of the optimized pattern [Figs. 2(c)and 2(d), right] with a random pattern and two naïve undersampling schemes of the full resolution image either through an unweighted averaging of 4×4 pixels [equivalent to the uniform pattern, Figs. 2(c) and 2(d), left] or a simple dropout [i.e., pick one pixel in every 4×4 and drop out the others, Fig. 2(b)]. All of the simulation groups employ an independently trained ISTA-Net for image reconstruction. In addition, we also used B-spline interpolation [28] to reconstruct the image that was undersampled through a simple dropout.

    Comparison of the reconstruction performance in the validation data set Set11 and BSD68, at an undersampling rate of 6.25%, through (a) a simple dropout, (b) an unweighted average (uniform pattern), (c) a random or an optimized illumination pattern (DeCIOPS) with a constraint of identical column, and (d) a random or an optimized illumination pattern (DeCIOPS). The PSNR and resolution of the reconstructed images are labeled below the exemplary sample. (e) PSNR of the reconstructed images of all 79 samples in the validation dataset for cases in (a)–(d). (f) Resolution of the reconstructed images of all 79 samples in the validation dataset for cases in (a)–(d). n.s., not significant; *, p<0.05; **, p<0.01; ***, p<0.001; ****, p<0.0001, in one-way analysis of variance (ANOVA).

    Figure 4.Comparison of the reconstruction performance in the validation data set Set11 and BSD68, at an undersampling rate of 6.25%, through (a) a simple dropout, (b) an unweighted average (uniform pattern), (c) a random or an optimized illumination pattern (DeCIOPS) with a constraint of identical column, and (d) a random or an optimized illumination pattern (DeCIOPS). The PSNR and resolution of the reconstructed images are labeled below the exemplary sample. (e) PSNR of the reconstructed images of all 79 samples in the validation dataset for cases in (a)–(d). (f) Resolution of the reconstructed images of all 79 samples in the validation dataset for cases in (a)–(d). n.s., not significant; *, p<0.05; **, p<0.01; ***, p<0.001; ****, p<0.0001, in one-way analysis of variance (ANOVA).

    4. EXPERIMENTAL RESULTS

    Experiment setup of DeCIOPS. The laser beam is spatially filtered to improve its spatial uniformity and symmetricity, collimated and expanded in size, and then incident onto a DMD. The beam is spatially modulated by the DMD and then shrunk in size by a 4f system formed by a tube lens and an objective lens. The light pattern is scanned by a resonant-galvo scanner set, where a resonant scanner and a galvanometer mirror are optically coupled through a relay lens set. The transmitted light from the sample is collected by a photodetector through a collection lens. The n×n pattern is generated by the DMD. With an additional 4f system with cylindrical lenses after the objective lens, the n×n pattern can be turned into n×1 size (Appendix A.1). The red dashed line (plane 1 and object plane) indicates the conjugate plane of the gray-scale pattern mask.

    Figure 5.Experiment setup of DeCIOPS. The laser beam is spatially filtered to improve its spatial uniformity and symmetricity, collimated and expanded in size, and then incident onto a DMD. The beam is spatially modulated by the DMD and then shrunk in size by a 4f system formed by a tube lens and an objective lens. The light pattern is scanned by a resonant-galvo scanner set, where a resonant scanner and a galvanometer mirror are optically coupled through a relay lens set. The transmitted light from the sample is collected by a photodetector through a collection lens. The n×n pattern is generated by the DMD. With an additional 4f system with cylindrical lenses after the objective lens, the n×n pattern can be turned into n×1 size (Appendix A.1). The red dashed line (plane 1 and object plane) indicates the conjugate plane of the gray-scale pattern mask.

    As the optical mode from the diode laser did not have uniform intensity, and the pattern could be corrupted by laser interference, we calibrated the DMD to ensure the illumination pattern on the sample plane matched well with the design (Appendix A.3). While our imaging system is naturally a CW light imaging system, we could also mimic the pulsed light source condition through an additional digital sampling step after the image acquisition (Appendix A.5).

    A. Reconstruction Results with a CW Light Source

    Comparison of the experimental results using different illumination patterns in the scanning in a CW illumination setting. (a)–(d) Experimental results of the sample: (a) butterfly, (b) cameraman, (c) house, and (d) the Flintstones. The different columns show the ground truth results using high-resolution point scanning, raw measurement using different illumination patterns at an undersampling rate of 6.25%, and the corresponding reconstruction results. (e) PSNR of the reconstructed images for a total of nine samples. (f) Spatial resolution of the reconstructed images for a total of nine samples, calculated from Fourier ring correlation. *, p<0.05; **, p<0.01; ***, p<0.001; ****, p<0.0001, in one-way ANOVA.

    Figure 6.Comparison of the experimental results using different illumination patterns in the scanning in a CW illumination setting. (a)–(d) Experimental results of the sample: (a) butterfly, (b) cameraman, (c) house, and (d) the Flintstones. The different columns show the ground truth results using high-resolution point scanning, raw measurement using different illumination patterns at an undersampling rate of 6.25%, and the corresponding reconstruction results. (e) PSNR of the reconstructed images for a total of nine samples. (f) Spatial resolution of the reconstructed images for a total of nine samples, calculated from Fourier ring correlation. *, p<0.05; **, p<0.01; ***, p<0.001; ****, p<0.0001, in one-way ANOVA.

    B. Reconstruction Results with a Low-Repetition-Rate Pulsed Light Source

    Comparison of the experimental results using different illumination patterns in the scanning in the low-repetition-rate pulsed light illumination setting. (a)–(d) Experimental results of the sample: (a) butterfly, (b) cameraman, (c) house, and (d) the Flintstones. The different columns show the ground truth results using high-resolution point scanning, raw measurement using different illumination patterns at an undersampling rate of 6.25%, and the corresponding reconstruction results. (e) PSNR of the reconstructed images for a total of nine samples. (f) Spatial resolution of the reconstructed images for a total of nine samples, calculated from Fourier ring correlation. *, p<0.05; **, p<0.01; ***, p<0.001; ****, p<0.0001, in one-way ANOVA.

    Figure 7.Comparison of the experimental results using different illumination patterns in the scanning in the low-repetition-rate pulsed light illumination setting. (a)–(d) Experimental results of the sample: (a) butterfly, (b) cameraman, (c) house, and (d) the Flintstones. The different columns show the ground truth results using high-resolution point scanning, raw measurement using different illumination patterns at an undersampling rate of 6.25%, and the corresponding reconstruction results. (e) PSNR of the reconstructed images for a total of nine samples. (f) Spatial resolution of the reconstructed images for a total of nine samples, calculated from Fourier ring correlation. *, p<0.05; **, p<0.01; ***, p<0.001; ****, p<0.0001, in one-way ANOVA.

    5. DISCUSSION

    A. Influence of SNR on Image Reconstruction

    (a) PSNR and (b) pixel resolution of the reconstructed images versus different SNRs in the raw measurement, for three different sampling patterns (CW configuration), performed through simulation at an undersampling rate of 6.25%. The results were averaged across nine samples used in the experiment and fitted with polynomial curves. (c) and (d) show the experimental results averaged across nine samples.

    Figure 8.(a) PSNR and (b) pixel resolution of the reconstructed images versus different SNRs in the raw measurement, for three different sampling patterns (CW configuration), performed through simulation at an undersampling rate of 6.25%. The results were averaged across nine samples used in the experiment and fitted with polynomial curves. (c) and (d) show the experimental results averaged across nine samples.

    B. Compressed Ratio and Size of the Optimized Pattern

    DeCIOPS reconstruction quality (a) PSNR and (b) pixel resolution dependence on the size of the optimized pattern, for an undersampling rate of 25% (2×2, red), 11.1% (3×3, green), 6.25% (4×4, blue), and 1.5625% (8×8, black), across all 79 samples in the validation dataset. Solid curve, mean; shaded area, standard deviation.

    Figure 9.DeCIOPS reconstruction quality (a) PSNR and (b) pixel resolution dependence on the size of the optimized pattern, for an undersampling rate of 25% (2×2, red), 11.1% (3×3, green), 6.25% (4×4, blue), and 1.5625% (8×8, black), across all 79 samples in the validation dataset. Solid curve, mean; shaded area, standard deviation.

    C. Optimized Pattern versus Random Pattern

    In the conventional switching-mask-based single-pixel camera, the random mask is one of the commonly used sampling bases, as it is incoherent with the spatial property of the sample. We found the same in the pattern-scanning scheme: the random pattern shows a superior performance compared with the uniform pattern. However, the optimized pattern, found by the auto-encoder through the end-to-end optimization, outperforms the random pattern. This is expected, as the random pattern attained by the Monte Carlo method is independent of the dataset of a specific task, unlike the optimized pattern. Furthermore, the performance improvement of the optimized pattern over the random pattern increases when the measurements are more highly undersampled, as seen in the comparison between 6.25% (4×4 pattern) and 1.5625% (8×8 pattern) (Table 1). This is attributed to more trainable parameters in the case of the optimized pattern and more uncertain random variables in the case of the random pattern.

    Comparison of the PSNR and Pixel Resolution across the Uniform Pattern, Random Pattern, and Optimized Pattern between 6.25% and 1.5625% Undersampling Rate in DeCIOPSa

    Undersampling RatePSNR (dB)Normalized Resolution (pixel)
    Uniform PatternRandom PatternOptimized PatternUniform PatternRandom PatternOptimized Pattern
    6.25%21.49±1.8321.82±1.7223.13±1.542.79±0.18  2.68±0.16  2.55±0.15
    1.5625%15.91±2.0116.78±1.9418.47±1.754.55±0.264.28±0.253.76±0.21  

    CW light configuration, SNR22, for all 79 samples in validation dataset.

    D. Comparison with Conventional Switching-Mask-Based Single-Pixel Camera

    Comparison of the reconstruction results between DeCIOPS and conventional switching-mask-based single-pixel camera. (a) The ground truth of an original object, butterfly. (b) Reconstruction result of DeCIOPS using ISTA-Net at an undersampling rate of 6.25%. (c) Reconstruction result of the switching-mask-based single-pixel camera imaging approach using ISTA-Net. Top row, simulation. Bottom row, experiment. The ground truth of the experiment is obtained by the high-resolution point scanning.

    Figure 10.Comparison of the reconstruction results between DeCIOPS and conventional switching-mask-based single-pixel camera. (a) The ground truth of an original object, butterfly. (b) Reconstruction result of DeCIOPS using ISTA-Net at an undersampling rate of 6.25%. (c) Reconstruction result of the switching-mask-based single-pixel camera imaging approach using ISTA-Net. Top row, simulation. Bottom row, experiment. The ground truth of the experiment is obtained by the high-resolution point scanning.

    E. Advantage of ISTA-Net in Image Reconstruction

    In DeCIOPS, we apply ISTA-Net as the decoder. ISTA-Net is a CS-induced neural network. Compared with conventional optimization algorithms where the regularization term is designed empirically, ISTA-Net is entirely data-driven and can learn optimized regularization through the neural network. Compared with other neural networks that could be used for super-resolution, such as U-Net [30,31] and densely connected super-resolution network (DCSRN) [32,33], the embedded CS algorithm in ISTA-Net fits better in the motivation of DeCIOPS and other compressed sensing framework, i.e., performing fewer measurements while being able to reconstruct high-resolution images. Indeed, when we compare the PSNR and pixel resolution of the reconstructed images across ISTA-Net, U-Net, and DCSRN, ISTA-Net shows the best performance (Appendix B). In addition to DeCIOPS, we believe ISTA-Net could also benefit other applications such as denoising [34], fast magnetic resonance imaging (MRI) [35], and other super-resolution imaging modalities [18].

    F. Advantage of End-to-End Optimized Auto-Encoder and Its Application in Future Imaging Systems

    In most existing optical imaging modalities, image formation is empirically designed and optimized, and the deconvolution or object reconstruction algorithm is subsequently tailored to the image formation process. The recent development of low-cost, advanced micro-optics manufacturing techniques, such as three-dimensional (3D) printing and micro–nano-fabrication [3640], allows rapid prototyping of user-designed optical elements, which opens new opportunities to redesign the image formation process that best fits the specific applications. Instead of sequentially designing the image formation and the reconstruction algorithm, their joint end-to-end optimization produces a global optimal solution [4143], which is the underlying principle of DeCIOPS. We use an auto-encoder to model the image formation and reconstruction within a single framework and perform end-to-end training to optimize the sampling pattern and ISTA-Net simultaneously. Our results show that the optimized sampling pattern indeed results in the best overall performance. Such an end-to-end training and data-driven approach prevents any empirical bias that may negatively impact the design. We envision that such an approach will enable many challenging applications such as super-resolution imaging [23,4246], 3D imaging [41,4749], and high-speed computational cameras [5052].

    G. Applicability in Two-Photon Microscopy

    While our imaging system used a CW light source, we mimicked the experimental condition of a pulsed light source and successfully demonstrated the applicability and excellent performance of DeCIOPS. This opens a new avenue to apply DeCIOPS in two-photon microscopy. In conventional two-photon microscopes [5,7,8], the image is acquired through pixel-by-pixel point scanning. While this enables deep tissue imaging as it resists light scattering, it reduces the imaging speed. Recently, there have been multiple reports applying compressed sensing in two-photon microscopy, with the same approach in the conventional switching-mask-based single-pixel camera [5355]. However, the improvement on the imaging speed is limited due to the low switching speed of DMDs or the liquid-crystal-based spatial light modulators. When applying DeCIOPS in two-photon microscopy, we expect that our approach will significantly increase the imaging speed and will notably benefit functional imaging through two-photon microscopy.

    H. Passive Light Illumination

    In our experiment, DeCIOPS is implemented using active light illumination (i.e., structured illumination), which is commonly used in biomedical imaging. In other imaging systems, passive light illumination may be preferred. In fact, any passive light illumination wide-field imaging using a focal plane array (i.e., camera) can be converted to DeCIOPS (Appendix C). A scanner can be added to the passive wide-field imaging to scan the entire image originally projected to the camera. By inserting a fixed mask with an appropriate aperture at the plane where the image is scanned, a single-pixel detector can measure the subsampled convolution between the mask and the original image. The object can then be reconstructed using the same algorithms in DeCIOPS for the active light illumination cases. This way, a structured detection version of DeCIOPS can be implemented.

    6. CONCLUSION

    We demonstrated a new high-speed imaging modality, DeCIOPS, by synthesizing the strength of conventional point scanning and single-pixel camera through compressed sensing. The high-speed imaging arises from the fast beam-scanning mechanism and a highly efficient sampling scheme through compressed sensing; meanwhile, an auto-encoder framework allows the simultaneous optimization of the image formation and reconstruction process in DeCIOPS. We validated DeCIOPS through both simulation and experiments in both CW and pulsed light source conditions. This new image modality can be adapted to any existing imaging systems using beam scanning, such as confocal microscope and two-photon microscope, or wide-field cameras with an added scanning system, and will benefit broad applications requiring high-speed imaging.

    Appendix A: Experimental Setup of DeCIOPS

    Optical Setup

    The optical setup of DeCIOPS illustrated in Fig.?5 scans an n×n pattern on the object and is suitable for low-repetition-rate pulsed light illumination settings. For CW light or high-repetition-rate pulsed light illumination, an n×1 pattern is scanned. When this pattern sweeps n columns, the information of n×n pixels can be integrated into a single measurement. To generate the n×1 pattern, we set the column to be identical in the n×n pattern and add a 4f system composed of cylindrical lenses after the objective lens. The n×n pattern is then shrunk in one dimension by a factor of n into the n×1 pattern (Fig.?11). The parameters of the lenses used in the setup are listed in Table?3.

    Experimental setup of DeCIOPS that generates an n×1 size pattern and scans it across the sample. The setup is similar to that generating the n×n size pattern shown in Fig. 5, but with a 4f system composed of cylindrical lenses added after the objective lens to shrink the original n×n size pattern in one dimension by a factor of n into the n×1 size. The red dashed line (plane 1 and object plane) indicates the conjugate plane of the gray-scale pattern mask.

    Figure 11.Experimental setup of DeCIOPS that generates an n×1 size pattern and scans it across the sample. The setup is similar to that generating the n×n size pattern shown in Fig. 5, but with a 4f system composed of cylindrical lenses added after the objective lens to shrink the original n×n size pattern in one dimension by a factor of n into the n×1 size. The red dashed line (plane 1 and object plane) indicates the conjugate plane of the gray-scale pattern mask.

    (a) Measured patterns on the sample (super-pixels) match well with the designed patterns. Each gray-scale super-pixel is generated by 32×32 binary pixels in the DMD. The left panel shows the cases for 4×4 patterns, and the right panel shows the cases for the 4×1 pattern. (b) A single spot pattern is generated for conventional point-scanning imaging to obtain the high-resolution ground truth of the sample. The spot size matches the size of a super-pixel. (c) Pixel-by-pixel comparisons between the measured patterns on the sample and the designed patterns show excellent matchings between the two.

    Figure 12.(a) Measured patterns on the sample (super-pixels) match well with the designed patterns. Each gray-scale super-pixel is generated by 32×32 binary pixels in the DMD. The left panel shows the cases for 4×4 patterns, and the right panel shows the cases for the 4×1 pattern. (b) A single spot pattern is generated for conventional point-scanning imaging to obtain the high-resolution ground truth of the sample. The spot size matches the size of a super-pixel. (c) Pixel-by-pixel comparisons between the measured patterns on the sample and the designed patterns show excellent matchings between the two.

    Measured patterns at the image plane stay consistent across different scanning angles.

    Figure 13.Measured patterns at the image plane stay consistent across different scanning angles.

    Data Acquisition

    Data acquisition is performed using a high-speed data acquisition card vDAQ and ScanImage software (Vidrio Technologies). As the illumination pattern continuously scans across each row, the data acquisition card samples the data from the photodetector at a rate higher than the single-pixel rate. The data acquired within the duration of a single pixel is then automatically averaged/integrated and saved as a single-pixel value. Compared with the full resolution single-point-scanning condition (256×256 pixels), we reduce the single-pixel rate and the number of scanning lines by 75% in DeCIOPS (CW light setting), reaching an undersampling rate of 6.25% (64×64 pixels of measurement).

    To mimic the low-repetition-rate pulsed light source condition, we sample the object with a high resolution at 2048×2048 pixels (corresponding to a high single-pixel rate) and then downsample the acquired image digitally into 64×64 pixels by dropping all of the other pixels. Here, each pixel has a small average/integration duration and could thus be considered as being acquired by a single light pulse.

    Estimation of Signal-to-Noise Ratio

    To calculate the SNR of the image in the experiment, we acquire the same image 20 times. For each pixel, we calculate the signal μ and the noise σ as the mean and the standard deviation across 20 measurements, respectively. The SNR for the pixel is then estimated as μ/σ. The SNR of the entire image is taken as the average of the SNR of all pixels.

    Appendix B: Comparison between ISTA-Net, U-Net, and DCSRN

    We compare the performance of ISTA-Net, U-Net, and DCSRN in object reconstruction. In the auto-encoder framework, the decoder implemented by ISTA-Net is replaced by U-Net or DCSRN. U-Net is widely used in image reconstruction and segmentation. It first condenses the size of the input images to extract its context and feature and then grows them in an expanding path to perform local reconstruction [30]. DCSRN, derived from densely connected convolutional networks [33], has a faster training speed and accurate reconstruction results and is commonly used in applications such as 2D or 3D biomedical super-resolution imaging. In addition to ISTA-Net, U-Net, and DCSRN, we used B-spline interpolation [28] to reconstruct the object undersampled through a simple dropout as a baseline. All of the simulation was completed on GPU RTX1080Ti 11?GB with 200 epochs and a batch size of five. In each decoder except for the B-spline, we learned an optimized illumination pattern. We used the validation data set to evaluate the PSNR and pixel resolution of the reconstructed objects. U-Net, DCSRN, and ISTA-Net all outperform B-spline interpolation. While U-Net and DCSRN do not show a significant difference in performance, ISTA-Net outperforms both U-Net and DCSRN with a 1.85??dB increase in PSNR and 6.55% improvement in resolution at an undersampling rate of 6.25% (Fig.?14). The simulation results demonstrated a clear advantage of ISTA-Net, which is a CS inspired neural network, in DeCIOPS.

    Comparison of (a) PSNR and (b) pixel resolution of the reconstructed objects of all 79 samples in the validation dataset for B-spline, U-Net, DCSRN, and ISTA-Net in the auto-encoder framework, at an undersampling rate of 6.25%. n.s., not significant; **, p<0.01; ***, p<0.001; ****, p<0.0001, in one-way ANOVA.

    Figure 14.Comparison of (a) PSNR and (b) pixel resolution of the reconstructed objects of all 79 samples in the validation dataset for B-spline, U-Net, DCSRN, and ISTA-Net in the auto-encoder framework, at an undersampling rate of 6.25%. n.s., not significant; **, p<0.01; ***, p<0.001; ****, p<0.0001, in one-way ANOVA.

    Appendix C: DeCIOPS Using Passive Light Illumination

    To implement DeCIOPS in a passive light illumination setting (i.e., structured detection), which is commonly used in photography, a scanner can be added to the passive wide-field imaging system to scan the entire image originally projected to the camera. By inserting a fixed mask with an appropriate aperture at the plane where the image is scanned, a single-pixel detector can measure the subsampled convolution between the mask and the original image (Fig.?15). The object can then be reconstructed using the same algorithms in DeCIOPS for the active light illumination cases.

    Optical setup of DeCIOPS with passive light illumination (i.e., structured detection) for applications such as photography.

    Figure 15.Optical setup of DeCIOPS with passive light illumination (i.e., structured detection) for applications such as photography.

    References

    [1] A. Rogalski. Infrared Detectors(2010).

    [2] M. F. Duarte, M. A. Davenport, D. Takhar, J. N. Laska, T. Sun, K. F. Kelly, R. G. Baraniuk. Single-pixel imaging via compressive sampling. IEEE Signal Process. Mag., 25, 83-91(2008).

    [3] M. P. Edgar, G. M. Gibson, M. J. Padgett. Principles and prospects for single-pixel imaging. Nat. Photonics, 13, 13-20(2019).

    [4] G. M. Gibson, S. D. Johnson, M. J. Padgett. Single-pixel imaging 12 years on: a review. Opt. Express, 28, 28190-28208(2020).

    [5] W. Denk, J. H. Strickler, W. W. Webb. Two-photon laser scanning fluorescence microscopy. Science, 248, 73-76(1990).

    [6] J. B. Pawley. Handbook of Biological Confocal Microscopy(2006).

    [7] W. R. Zipfel, R. M. Williams, W. W. Webb. Nonlinear magic: multiphoton microscopy in the biosciences. Nat. Biotechnol., 21, 1369-1377(2003).

    [8] F. Helmchen, W. Denk. Deep tissue two-photon microscopy. Nat. Methods, 2, 932-940(2005).

    [9] L. Beiser. Fundamental architecture of optical scanning systems. Appl. Opt., 34, 7307-7317(1995).

    [10] E. J. Candès. Compressive sampling. Proceedings of the International Congress of Mathematicians, 1433-1452(2006).

    [11] E. Candes, J. Romberg. Sparsity and incoherence in compressive sampling. Inverse Prob., 23, 969(2007).

    [12] E. J. Candès, M. B. Wakin. An introduction to compressive sampling. IEEE Signal Process. Mag., 25, 21-30(2008).

    [13] A. P. Spencer, B. Spokoyny, S. Ray, F. Sarvari, E. Harel. Mapping multidimensional electronic structure and ultrafast dynamics with single-element detection and compressive sensing. Nat. Commun., 7, 10434(2016).

    [14] L. Bian, J. Suo, Q. Dai, F. Chen. Experimental comparison of single-pixel imaging algorithms. J. Opt. Soc. Am. A, 35, 78-87(2018).

    [15] J. B. Sampsell. Digital micromirror device and its application to projection displays. J. Vac. Sci. Technol. B, 12, 3242-3246(1994).

    [16] Z. H. Xu, W. Chen, J. Penuelas, M. Padgett, M. J. Sun. 1000  fps computational ghost imaging using LED-based structured illumination. Opt. Express, 26, 2427-2434(2018).

    [17] H. Wu, Z. Zheng, Y. Li, W. Dai, H. Xiong. Compressed sensing via a deep convolutional auto-encoder. IEEE Visual Communications and Image Processing (VCIP), 1-4(2018).

    [18] J. Zhang, B. Ghanem. ISTA-Net: interpretable optimization-inspired deep network for image compressive sensing. IEEE Conference on Computer Vision and Pattern Recognition, 1828-1837(2018).

    [19] A. Beck, M. Teboulle. A fast iterative shrinkage-thresholding algorithm for linear inverse problems. SIAM J. Imag. Sci., 2, 183-202(2009).

    [20] N. Pavillon, N. I. Smith. Compressed sensing laser scanning microscopy. Opt. Express, 24, 30038-30052(2016).

    [21] L. Fang, F. Monroe, S. W. Novak, L. Kirk, C. R. Schiavon, S. B. Yu, T. Zhang, M. Wu, K. Kastner, Y. Kubota, Z. Zhang, G. Pekkurnaz, J. Mendenhall, K. Harris, J. Howard, U. Manor. Deep learning-based point-scanning super-resolution imaging(2019).

    [22] B. Li, Q.-R. Yan, Y.-F. Wang, Y.-B. Yang, Y.-H. Wang. A binary sampling Res2net reconstruction network for single-pixel imaging. Rev. Sci. Instrum., 91, 033709(2020).

    [23] H. Wang, Y. Rivenson, Y. Jin, Z. Wei, R. Gao, H. Gunaydin, L. A. Bentolila, C. Kural, A. Ozcan. Deep learning enables cross-modality super-resolution in fluorescence microscopy. Nat. Methods, 16, 103-110(2019).

    [24] J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, L. Fei-Fei. Imagenet: a large-scale hierarchical image database. IEEE Conference on Computer Vision and Pattern Recognition, 248-255(2009).

    [25] K. Kulkarni, S. Lohit, P. Turaga, R. Kerviche, A. Ashok. Reconnet: non-iterative reconstruction of images from compressively sensed measurements. IEEE Conference on Computer Vision and Pattern Recognition, 449-458(2016).

    [26] D. Martin, C. Fowlkes, D. Tal, J. Malik. A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics. 8th IEEE International Conference on Computer Vision (ICCV), 416-423(2001).

    [27] S. Koho, G. Tortarolo, M. Castello, T. Deguchi, A. Diaspro, G. Vicidomini. Fourier ring correlation simplifies image restoration in fluorescence microscopy. Nat. Commun., 10, 3103(2019).

    [28] T. M. Lehmann, C. Gonner, K. Spitzer. Addendum: B-spline interpolation in medical image processing. IEEE Trans. Med. Imaging, 20, 660-665(2001).

    [29] T. A. Pologruto, B. L. Sabatini, K. Svoboda. ScanImage: flexible software for operating laser scanning microscopes. Biomed. Eng. Online, 2, 13(2003).

    [30] O. Ronneberger, P. Fischer, T. Brox. U-net: convolutional networks for biomedical image segmentation. International Conference on Medical Image Computing and Computer-Assisted Intervention, 234-241(2015).

    [31] J. Feng, J. Deng, Z. Li, Z. Sun, H. Dou, K. Jia. End-to-end Res-Unet based reconstruction algorithm for photoacoustic imaging. Biomed. Opt. Express, 11, 5321-5340(2020).

    [32] Y. Chen, Y. Xie, Z. Zhou, F. Shi, A. G. Christodoulou, D. Li. Brain MRI super resolution using 3D deep densely connected neural networks. IEEE 15th International Symposium on Biomedical Imaging (ISBI), 739-742(2018).

    [33] G. Huang, Z. Liu, L. Van Der Maaten, K. Q. Weinberger. Densely connected convolutional networks. IEEE Conference on Computer Vision and Pattern Recognition, 4700-4708(2017).

    [34] B. Lee, B. Ku, W.-J. Kim, S. Kim, H. Ko. Denoising ISTA-Net: learning based compressive sensing with reinforced non-linearity for side scan sonar image denoising. J. Acoust. Soc. Korea, 39, 246-254(2020).

    [35] E. D. W. N. Pezzotti, S. Yousefi, M. S. Elmahdy, J. van Gemert, C. Schülke, M. Doneva, T. Nielsen, S. Kastryulin, B. P. F. Lelieveldt, M. J. P. van Osch, M. Staring. Adaptive-CS-Net: FastMRI with adaptive intelligence(2019).

    [36] X. Q. Zhou, Y. H. Hou, J. Q. Lin. A review on the processing accuracy of two-photon polymerization. AIP Adv., 5, 030701(2015).

    [37] T. Gissibl, S. Thiele, A. Herkommer, H. Giessen. Two-photon direct laser writing of ultracompact multi-lens objectives. Nat. Photonics, 10, 554-560(2016).

    [38] J. W. Li, P. Fejes, D. Lorenser, B. C. Quirk, P. B. Noble, R. W. Kirk, A. Orth, F. M. Wood, B. C. Gibson, D. D. Sampson, R. A. McLaughlin. Two-photon polymerisation 3D printed freeform micro-optics for optical coherence tomography fibre probes. Sci. Rep., 8, 14789(2018).

    [39] N. Mohammad, M. Meem, X. Wan, R. Menon. Full-color, large area, transmissive holograms enabled by multi-level diffractive optics. Sci. Rep., 7, 5789(2017).

    [40] N. Vaidya, O. Solgaard. 3D printed optics with nanometer scale surface roughness. Microsys. Nanoeng., 4, 18(2018).

    [41] Y. C. Wu, V. Boominathan, H. J. Chen, A. Sankaranarayanan, A. Veeraraghavan. PhaseCam3D-learning phase masks for passive single view depth estimation. IEEE International Conference on Computational Photography(2019).

    [42] V. Sitzmann, S. Diamond, Y. F. Peng, X. Dun, S. Boyd, W. Heidrich, F. Heide, G. Wetzstein. End-to-end optimization of optics and image processing for achromatic extended depth of field and super-resolution imaging. ACM Trans. Graph., 37, 114(2018).

    [43] Q. L. Sun, J. Zhang, X. Dun, B. Ghanem, Y. F. Peng, W. Heidrich. End-to-end learned, optically coded super-resolution SPAD camera. ACM Trans. Graph., 39, 9(2020).

    [44] Y. Li, B. Sixou, F. Peyrin. A review of the deep learning methods for medical images super resolution problems. IRBM(2020).

    [45] W. Ouyang, A. Aristov, M. Lelek, X. Hao, C. Zimmer. Deep learning massively accelerates super-resolution localization microscopy. Nat. Biotechnol., 36, 460-468(2018).

    [46] M. Weigert, U. Schmidt, T. Boothe, A. Muller, A. Dibrov, A. Jain, B. Wilhelm, D. Schmidt, C. Broaddus, S. Culley, M. Rocha-Martins, F. Segovia-Miranda, C. Norden, R. Henriques, M. Zerial, M. Solimena, J. Rink, P. Tomancak, L. Royer, F. Jug, E. W. Myers. Content-aware image restoration: pushing the limits of fluorescence microscopy. Nat. Methods, 15, 1090-1097(2018).

    [47] S. Zhang. High-speed 3D shape measurement with structured light methods: a review. Opt. Laser Eng., 106, 119-131(2018).

    [48] Y. C. Wu, Y. Rivenson, H. D. Wang, Y. L. Luo, E. Ben-David, L. A. Bentolila, C. Pritz, A. Ozcan. Three-dimensional virtual refocusing of fluorescence microscopy images using deep learning. Nat. Methods, 16, 1323-1331(2019).

    [49] H. Pinkard, Z. Phillips, A. Babakhani, D. A. Fletcher, L. Waller. Deep learning for single-shot autofocus microscopy. Optica, 6, 794-797(2019).

    [50] Y. Ma, Y. Lee, C. Best-Popescu, L. Gao. High-speed compressed-sensing fluorescence lifetime imaging microscopy of live cells. Proc. Natl. Acad. Sci. USA, 118, e2004176118(2020).

    [51] D. Qi, S. Zhang, C. Yang, Y. He, F. Cao, J. Yao, P. Ding, L. Gao, T. Jia, J. Liang. Single-shot compressed ultrafast photography: a review. Adv. Photon., 2, 014003(2020).

    [52] L. Gao, J. Y. Liang, C. Y. Li, L. H. V. Wang. Single-shot compressed ultrafast photography at one hundred billion frames per second. Nature, 516, 74-77(2014).

    [53] M. Alemohammad, J. Shin, D. N. Tran, J. R. Stroud, S. P. Chin, T. D. Tran, M. A. Foster. Widefield compressive multiphoton microscopy. Opt. Lett., 43, 2989-2992(2018).

    [54] P. Wijesinghe, A. Escobet-Montalban, M. Z. Chen, P. R. T. Munro, K. Dholakia. Optimal compressive multiphoton imaging at depth using single-pixel detection. Opt. Lett., 44, 4981-4984(2019).

    [55] Y.-X. Ren, C. Kong, H. He, X. Zeng, K. K. Tsia, K. K. Wong. Encrypted wide-field two-photon microscopy with single-pixel detection and compressed sensing. Appl. Phys. Express, 13, 032007(2020).

    [56] J. N. Stirman, I. T. Smith, M. W. Kudenov, S. L. Smith. Wide field-of-view, multi-region, two-photon imaging of neuronal activity in the mammalian brain. Nat. Biotechnol., 34, 857-862(2016).

    Copy Citation Text
    Kangning Zhang, Junjie Hu, Weijian Yang. Deep compressed imaging via optimized pattern scanning[J]. Photonics Research, 2021, 9(3): 03000B57
    Download Citation
    Special Issue: DEEP LEARNING IN PHOTONICS
    Received: Oct. 7, 2020
    Accepted: Jan. 13, 2021
    Posted: Jan. 15, 2021
    Published Online: Mar. 2, 2021
    The Author Email: Weijian Yang (wejyang@ucdavis.edu)