This PDF file contains the front matter associated with SPIE Proceedings Volume 11896 including the Title Page, Copyright information, and Table of Contents.

Computational imaging is a powerful imaging framework by cooperating optics and information science. In this talk, I will present our research activities related to computational imaging with scattering media and machine learning.

Compressed ultrafast photography (CUP) is the fastest receive-only single-shot imaging technique up to now. By combining compressed sensing and streak imaging, CUP is able to capture ultrafast dynamics in a single shot. As a powerful tool for researching ultrafast phenomena, it has been widely applied in lots of areas. To meet the demand for more precise dynamics information and higher dimension in some application, many improvements have been conduct in CUP. For example, we have raised total variation-block match 3D filter algorithm and augmented Lagrange-deep learning hybrid algorithm to improve the reconstructed image quality of CUP, and set up a stereo-volumetric CUP system to capture 5 dimension dynamic information in a single shot. Besides, we have also developed another single-shot ultrafast optical imaging technique, chirped spectral mapping ultrafast photography (CSMUP), which utilized the spectral-temporal mapping to exact temporal information from hyperspectral image.

Transport of intensity equation (TIE) is an established quantitative phase imaging (QPI) method derived from paraxial approximation, however, this approximation limits TIE to retrieve high-resolution QPI. Thus, we present a dark-field transport of intensity equation (DFTIE) approach to improve resolution, solving the attenuation and blurring at high frequencies, which is combined iterative TIE phase retrieval with high-angle illumination coherent-mode decomposition to improve the maximum resolution under low numerical aperture objective. Simulation results demonstrate that DFTIE can achieve high-contrast and high-resolution QPI over the whole theoretical bandwidth, showing efficiency for high-throughput imaging. With the simple mode (only bright-field and dark-field are required) and multiplexing illumination, the imaging signal-to-noise ratio is higher, offering a flexible and cost-effective alternative for biomedical research and cellular investigations.

A new multimode fiber micro-endoscopy that integrates fiber calibration and imaging phase is proposed. The approach utilizes the back-scattered patterns emitted by the fluorescence targets at the fiber distal end to reconstruct the reflection matrix by phase retrieval algorithm. Simulations show that selective focusing through the fiber is directly achieved using the eigenvectors decomposed from the reflection matrix. It is possible to reconstruct the full fluorescent objects based on speckle correlations. Further experiments will be conducted to explore the focusing and imaging performance. The fluorescence-based reflection matrix method has the potential for re-calibration in situ and robust imaging under moderate perturbations, promising for practical micro-endoscopic applications.

To improve the focus tunability and reduce the difficulties in fabrication, we present a novel optofluidic varifocal lens actuated by dielectric elastomer sandwiched by two conductive liquids with different refractive indexes. The lens has a compact structure: two grid frames, two passive membranes, a dielectric elastomer actuator membrane, and two conductive liquids. Two cavities of the varifocal lens are filled with different conductive liquids. The conductive liquids are not only employed as the material of the optofluidic varifocal lens because of their high transparency but also work as the compliant electrode of the dielectric elastomer actuator membrane due to their conductibility. The results experiments show that the focal length variation of the lens is greater than 100%, the response time is approximately 75 ms. The lens is free from the compliant electrodeswhich greatly reduces the difficulties in fabrication.

We demonstrate a snap-shot ghost diffraction imaging approach with potential features of simultaneous recovery of amplitude and phase of a complex-valued object from a single-shot recording of the fields at the detectors. The technique utilizes the spatial averaging as an effective replacement of ensemble averaging in the execution of the cross-correlation of intensity fluctuations at the detector plane. Furthermore, the approach adopts the concept of holography in combination with the ghost diffraction scheme for the simultaneous recovery of phase distribution along with the amplitude of the object. The proposed method is expected to find applications in the two- and three-dimensional real-time quantitative imaging, biological microscopy, tomography, and super-resolution imaging, etc.

The development of tumor is closely related to extracellular matrix, which changes the biomechanical behavior of cells.Research have prepared polyacrylamide hydrogel substrates of differing stiffness according to the hardness values of breast tissue under normal and tumor physiological conditions. Then AFM was used to measure the mechanical properties of breast cells with different degrees of malignancy grown on different stiffness substrates. To explore the reasons for the changes in the young’s modulus of three breast cells, the distribution of cellular actin filaments were observed with a confocal microscope. These results showed that when the substrate hardened, the viscoelasticity of benign breast cells increased significantly, and the other two cancer cells also changed to some extent. We also found that the harder the substrate, the more conducive to the spreading behavior of cells, and the weaker response of malignant cells to substrates.

The continuous zoom multi configuration method is developed, and the correlative model is established, which overcomes the limitations of traditional methods, and realizes the expected system evaluation criteria，such as MTF, REA, RMS, etc..through the balance between value function, optimization algorithm and model parameterization.

In this paper, a 2-D sizing measuring system of regular particles is developed for characterizing the shapes by simultaneously measuring their length and width. Due to the multi-regulars shape of particles, it is usually required in situ 2D semi-automatic imaging analysis technique to describe their different shapes. In our works, edges of particles in binary image are first detected based on automatic threshold decomposition of the original gray-scale image. Then, the internal holes are filled in some individual detection region. In order to separate from different regular shapes, a selected threshold for the rectangularity filter has been applied. Then, by use of Euclidean distance map (EDM), the size measurement of individual particle is calculated. Finally, a series of experiments on these selected electron micro graphs , which contains particle with sizes from 10nm to 200nm, are respectively carried out to verify the performance of previous image analysis technology on our developed software. Those results are most promising for on-machine applications in naon-dimensional measurement of regular particles.

Light sheet fluorescence microscopy (LSFM) utilizes a light sheet to optically section samples, giving it the advantages of high signal-to-noise ratio, low phototoxicity and rapid three-dimensional imaging, which is an ideal tool for long-time microscopic observation of living samples. However, due to the inherent characteristics of Gaussian light sheet, there is a contradiction between the field of view and depth resolution of LSFM. In life science research, high depth resolution is necessary to resolve fine structures of biological systems such as neurons. However, in order to maintain a large field of view, the thickness of Gaussian light sheet is limited. Here, we propose a depth resolution enhancement method for LSFM. The incorporation of light field imaging into LSFM allows the acquirement of light field information within the Gaussian light sheet. By using the light field back projection (LFBP) reconstruction algorithm and depth estimation algorithm, it is possible to obtain the depth information in a light sheet, so as to improve the depth resolution. We built a light field light sheet fluorescence microscope (LFLSFM), and demonstrated it by fluorescent microsphere experiments. The proposed method can achieve a depth resolution of 0.5um when the thickness of the light sheet is kept at 6um to ensure a large field of view. Therefore, the method has potentialities in visualizing neural network of tissues and organs such as mouse brain.

The stacked autoencoder (SAE) neural network applied to diffuse optical tomography (DOT) achieves accurate and stable detection of the position and size of tissue abnormality. The quality of modeling data influences the robustness and the accuracy of the model, the measurability of the model determines the effective range of the data cleaning method used in clinical practice. In order to determine the effective range of this method in clinical use, we analyze the measurability of anomaly detection based on DOT method. The analysis result is used as a priori information to clean the neural network sample data set used in this work. The results show that excluding the data outside the measurable range, the proposed method enables the network to achieve a prediction accuracy of 99% within the measurable range and achieves rapid and accurate detection of the position and size of abnormality in the tissue.

A common polarization measurement system consists of a quarter-wave plate and a linear polarizer, and the rotation of the wave plate can provide different phase retardance. However, the rotating phase retarder will introduce some unwanted frequency components in the light intensity signal, which may affect the correct information extraction for dynamic observation of the optical process. In this work, we analyzed the effective working frequency band of the polarization measurement system and pointed out how to judge the reliability of dynamic periodic polarization signals. We present the parameter selection strategy of the cutoff frequency of the filter, sampling frequency and rotation angular velocity for a specific dynamic polarization change of scattered light. Finally, we apply our work in the polarization monitoring of tissue optical clearing and show the improvement of the measurement stability and accuracy.

For specular reflection surface detection, high reflectivity is a large challenge to effectively extract the depth information of surface. Phase Measuring Deflectometry (PMD) based three-dimensional shape measurement is proposed for solving this problem. In this study, PMD method based on the characteristic of high specular reflectivity is used to perform structured light imaging on the glass surface to obtain depth information on the surface of the glass panel. In this paper, we propose a new image reconstruction method suitable for imaging specular reflection surface defects. According to the characteristic of the glass panel, the proposed method has a phase pre-unwrapping process and improves the least square method of unfolding and folding the phase algorithm. The experimental results show that the proposed method is more robust for imaging and detection of high-reflective plane than the traditional least squares method.

The wide application of the image super-resolution algorithms significantly improves the visual quality of infrared images. In this paper, an infrared image super-resolution reconstruction method based on a closed-loop regression network is proposed. The residual channel attention block is introduced into the up-sampling module group, which effectively improves the capacity of the network and increases the number of feature maps, enhances the extraction and recovery ability of infrared image features, and is conducive to the recovery of image details. Compared with other infrared information recovery methods previously proposed, the proposed method has obvious advantages in high-resolution detail acquisition capability. Neural network through closed-loop regression, this scheme overcomes the LR image to HR image defects in nonlinear mapping function, by introducing additional constraints on the LR data to reduce the space of the possible functions.

Photon-counting imaging detectors based on microchannel plate (MCP) and position-sensitive anode are characterized by extremely high sensitivity, as well as the ability to detect single photons. In this study, a visible light photon-counting imaging detector based on induction readout was developed for the detection of weak light intensities, such as photon counting radar and biofluorescence lifetime imaging. The design is based on a 25 mm diameter multi-alkali S20 photocathode followed by a MCP stack, and read out by a high-resistance Ge layer anode. The position-sensitive anode was located at the atmosphere side of the Ge substrate. Such a detector was advantageous in terms of reconfigurability and detachability. The imaging performance of the detector was tested by using the wedge and strip anode to decode the photon event position information. The experimental results showed a detector gain reaching about 3×10⁶ , pulse height resolution (PHR) of about 105%, and spatial resolution better than 100μm.

For the distributed aperture synthesis imaging system based on digital holography, the influence of sub-aperture’s tilt and displacement on the synthetic aperture image is analyzed firstly, and then a correction method based on the image correlation and image sharpness optimization is proposed and validated by experiments. The tilt of the photo detector in the sub aperture leads to the displacement of the reconstructed target amplitude while the displacement leads to the piston, tilt and displacement of the reconstructed target amplitude. For the two errors, the image correlation method can be used to estimate the tilt and displacement values initially, and then the tilt and displacement values are iterated by the stochastic parallel gradient descent algorithm to improve the sharpness of synthetic aperture images. The experimental results show that the sharpness of synthetic aperture image are significantly improved after correction, but speckle noise has an obvious influence on the correction accuracy.

Compressed sensing(CS) technology can efficiently restore information from far fewer measurements than what Nyquist sampling theory requires. Currently, most CS reconstruction algorithms only reconstruct objects from spacial or temporal compressive measurements. Given the complexity and the difficulty, even using neural networks, it is difficult to reconstruct an object form spatial-temporal compressive measurements. In this paper, we represents the imaging process in spatial-temporal compressive imaging (STCI) into a cascaded process of spatial compressive imaging(SCI) followed by temporal compressive imaging(TCI). Thus to reconstruct an object from STCI, we first reconstruct multiple object frames from a single STCI measurement frame, and then improve object frames’ resolution. The TCI reconstruction algorithm used in this paper is TwIST algorithm. To improve object frame spatial resolution, we use a deep learning network SRResNet+. Besides improving resolution, SRResNet+ can also suppress residual error in TCI reconstruction frames. We verify our idea using numerical experiments. When the compressive ratio for STCI is 16:1, or the compressive ratios for SCI and TCI both are 4:1, the reconstructions obtained using TwIST followed by SRResNet+ present a PSNR value 29dB.

In the dual-camera phase retrieval method, the phase is solved by positive- and negative- defocusing images obtained through a single exposure after dual cameras are installed on an inverted microscope. However, due to the installation error of the cameras, translation and rotation of images exist between the images, resulting inaccurate phase retrieval results. In this paper, we proposed a dual-camera phase retrieval method based on fast adaption image restoration and transport of intensity equation. Firstly, let the positive-defocusing image be the reference image. Then using the fast adaption image restoration algorithm to find the texture information in order to find best matching block quickly. According to the number of high frequency information of the block, the size of block can be defined in order to increase the precision and speed of the restoration. After that, priority can be change as the sum of two parts, which can avoid the situation of 0 priority. Then, burring the boundary point of restored image in order to reduce the block effect. Finally, the transport of intensity equation can be used in phase retrieval results. Comparing with the normal algorithm, this method can restore the image much better.