This PDF file contains the front matter associated with SPIE Proceedings Volume 12766, including the Title Page, Copyright information, Table of Contents, and Conference Committee information.

Femtosecond-laser-driven light sources on a water film have been applied for computational broadband imaging. The spatially selective generated light sources on a water film form a structured illumination pattern with a broadband wavelength including visible and non-visible region. The spatial position of the light source was controlled by beam deflection using a galvanometer scanner or phase modulation with computer-generated holograms displayed on a liquid crystal on silicon spatial light modulator. The light source excited on a water film can produce a broad emission spectrum that includes x-rays and terahertz waves in addition to the visible region. We have demonstrated x-ray and visible imaging by using the femtosecond-laser-driven light sources which was two-dimensionally generated on a water film. Furthermore, the imaging time was reduced while maintaining the number of pixels in the reconstructed image by using compressed sensing algorithms and coded illumination patterns.

It is a challenge for conventional monocular-camera single-light source eye tracking methods to achieve high-speed eye tracking. Human gaze motion is a high-speed and miniature eye movement. Eye tracking requires a high-speed sampling frequency. In this work, an eye tracking method was proposed to overcomes the above limitation. The dual-ring infrared lighting source was designed to achieve bright and dark pupils in high-speed. The eye tracking method used a dual-ring infrared lighting source and synchronized triggers for the even and odd camera frames to capture bright and dark pupils. A pupillary corneal reflex was calculated by the center coordinates of the Purkinje spot and the pupil. A map function was established to map the relationship between pupillary corneal reflex and gaze spots. The gaze coordinate was calculated based on the mapping function. The detection time of each frame was less than five milliseconds, which achieved the purpose of high-speed eye tracking of the human gaze.

Data in single-molecule localization microscopy (SMLM) contain a large amount of biological information, and accurate quantitative analysis of these data is crucial for studying cellular functions at the biomolecular level. Current SMLM analysis tools often rely on a single method and do not fully consider the potential effects of imaging artifacts on the accuracy of analysis. Here we developed an easy-to-use ImageJ plugin called DecodeSTORM, which integrates multiple quantitative analysis methods (including segmentation, clustering, spatial statistics and co-localization), and also provides various artifact correction methods (including drift correction and localization filtering). Users are free to combine these methods as needed to improve the accuracy of quantitative analysis. DecodeSTORM aims to provide an easy data analysis tool for biological users who are looking for a more accurate data analysis in SMLM.

By acquiring three-dimensional profiles of biological tissues, interventions can be performed with increased speed and accuracy, driving the development of next-generation image guided therapy. However, current three-dimensional reconstruction techniques relying on feature detection and matching struggle with tissues lacking distinct features, resulting in relatively sparse reconstruction results. In this paper, we propose a data-driven method for reconstructing three-dimensional surfaces from a single polarimetric image, utilizing physics-based priors. We constructed a calibrated imaging system consisting of a polarization camera and a 3D scanner to collect polarization information and ground truth 3D data. Using this system, we created a dataset with organ models, capturing polarization images, depth maps, and surface normal maps under different lighting conditions. To achieve our goal, we designed a deep neural network based on the Unet architecture. This network takes the polarization image and prior physical parameter maps (phase angle, degree of polarization, and unpolarized intensity) as inputs and is trained to output the surface normal map and relative depth map of the organ. Experimental results on the tissue phantom dataset demonstrate the effectiveness of our method in generating dense reconstruction results, even for the regions lacking distinct features. Furthermore, we validated the robustness of our method to changes in the light source direction, showcasing its ability to handle variations in lighting conditions. Overall, our proposed data-driven approach provides a promising solution for dense three-dimensional reconstruction from a single polarimetric image, leveraging physics-based priors and deep learning techniques.

To overcome the ill-conditioning of the NIR fluorescence molecular tomography (FMT) inverse problem, neural networks are commonly used for reconstruction to improve the accuracy and reliability of imaging. This paper aims to investigate the impact of different neural network structures on the reconstruction performance of FMT for improved effect. In this study, the finite element solution of the Laplace-transformed time-domain coupled diffusion equation serves as the forward model for FMT, an improved stacked autoencoder (SAE) network is used and applied to FMT reconstruction. In the study, the SAE was set as a four layers network model structure, of which two layers were used for the hidden layer of the network. When the number of neurons in hidden layer 1 is smaller than hidden layer 2, the network is referred to as a decreasing network structure, and vice versa for an increasing network structure. The input data to the network consists of surface fluorescence intensity values collected by detectors around the heterogeneity. The output data of the network consists of fluorescence intensity values on partitioned nodes obtained through finite element method (FEM) partitioning. The experimental results demonstrate that the increasing network structure exhibits better imaging accuracy, fewer artifacts, and a more stable network model in FMT reconstruction. Through this study of the impact of SAE network architecture on FMT reconstruction, we have identified the optimal network model, which holds significant guidance for the application of neural networks in the field of FMT.

Ghost imaging also known as correlated imaging, is an indirect imaging method that reconstructs target object information through non-local intensity correlation. Traditional ghost imaging uses laser propagated to a rotating ground glass to generate speckle, The transmitted light passing through the object is then detected by a bucket detector, and the object image is reconstructed through non-local intensity correlation. However, the presence of speckle causes irregular noise points in the reconstructed images, leading to low contrast and signal-to-noise ratio. Aiming at the shortcoming of ghost imaging, suppressing the speckle noise via transform domain decomposition and reconstruction is proposed in this paper for high quality ghost imaging. The reconstructed image quality was quantitatively evaluated using peak signal-to-noise ratio (PSNR), Structural Similarity Index (SSIM), and mutual information(MI). Compared to existing approaches, the proposed method provides significant improvements to optimize image quality.

As a key and unique feature, anisotropy is the variability of material properties in different directions caused by molecular conformation and structural arrangement. The optical properties of anisotropic materials include birefringence and dichroism, and the characterization of these properties can be used to detect changes in the microstructure of materials. Conventional polarized optics imaging and photoacoustic imaging are based on scattering and contact measurements respectively, limiting many applications. Photoacoustic remote sensing (PARS) is an all-optical, non-contact photoacoustic measurement technique that provides specific light absorption contrast without coupling or labeling. However, PARS usually ignores tissue uptake anisotropy, which may result in missing some unique tissue properties and information. To obtain more dimensional information on targets, here we propose polarized photoacoustic remote sensing (P-PARS) microscopy. The system uses linearly polarized light in different directions as the excitation source and indicates the anisotropic optical absorption of the target by detecting changes in the reflected intensity of the interrogation light. It can simultaneously provide images of optical absorption contrast and the degree of anisotropy. Experimental results of testing materials and biological tissues demonstrated the feasibility and stability of the P-PARS. This method provides a new noncontact label-free strategy for anisotropy detection, prefiguring important potential for anisotropic material inspection and biological tissue imaging.

Optical microscopy is an indispensable imaging tool in life science. However, light cannot be focused owing to high scattering, and hence, the imaging depth and spatial resolution are restricted. Here, we propose an imaging method that combines wavefront shaping and image scanning microscopy. The reflected signal is used as feedback to acquire an optimal phase that can refocus the scattered light behind the scattering media. The experimental results show that the proposed method works in multilayer scattering media and can improve both the resolution and imaging depth of optical microscopy.

Phase diversity technique (PD) can jointly estimate the wavefront aberration and the target image of an optical imaging system. The PD technique reconstructs images by acquiring a focal plane image of optical system and one or more images with known aberrations (often selected defocus). Due to the simple construction of the optical system, the ability to detect discontinuous co-phase errors, and its applicability to both point sources and extended targets, The PD technique is uniquely suited for spatial target imaging applications, especially for the detection of multi-aperture piston errors. However, in a spatially low-illumination environment, Poisson noise as the main noise source of the imaging system seriously affects the accuracy of the reconstructed images. In this paper, we propose a method of phase diversity technique based on a fast Non-local Means (NLM) algorithm for reconstructing single-aperture images or multi-aperture images. For the two cases of single-aperture imaging and multi-aperture imaging with piston errors in spatial low illumination conditions, the method is used to solve the sensitivity problem of Poisson noise during image reconstruction. Numerical simulation results show that our method has significant improvement in structural similarity of the recovered images compared with the traditional phase diversity technique, and also is faster than the common non-local mean algorithm. The combination of this fast non-local means algorithm which using integral images and the phase diversity technique greatly reduce the computation time. The field experimental results and simulation results show good agreement. The new method would be useful in the AO system with active Poisson noise.

Frequency-shift super-resolution (SR) microscopy, such as structured illumination microscopy (SIM) and Fourier ptychographic microscopy (FPM), can break the diffraction limit for the imaging of both fluorescently labeled and label-free samples by transferring the high spatial-frequency information into the passband of microscope. However, conventional SIM microscopy systems tend to be bulky and expensive, which limits its applications in various fields. Therefore, we’ve developed some chip-based frequency-shift SR technologies, which is compatible with conventional microscope, and can be further designed for portable imaging systems such as smart phone and so on. We also developed a deep-learning based imaging method to improve the imaging speed of frequency-shift super-resolution microscopy, which enables low-cost and fast super-resolution imaging for real-time live cell biological studies.

According to different scenarios and needs, the rotating imaging optical satellite designs three task observation modes such as wide-area coverage, encrypted observation and regional video, the long edge of the camera field of view is placed radially along the turntable, the short edge of the field of view is placed tangentially, the camera pendulum axis is placed along the short side of the field of view, the pendulum mirror can be adjusted in three gears, the satellite adopts nadir pointing, that is, the rotation axis of the turntable is in the direction of the nadiel, the turntable maintains a normal working speed of 24 ° / s, rotates every 15s, and adjusts one gear per rotation of the camera pendulum mirror. After rotating 3 times in 50s, it can cover the range of 5~60° in the direction of nadir direction. The image obtained based on the above mode has rotation misalignment, different angles, abnormal speed imaging, integration time jump, and internal distortion in a single frame image. Therefore, how to perform sensor correction for multi-angle rotating images to achieve continuous, complete and distortion-free restoration is a key scientific problem to be solved in this paper. The purpose of optical satellite sensor correction is to design the imaging payload and imaging mode, realize seamless image stitching, eliminate image distortion caused by lens distortion, CCD distortion, integrated time jump, allospeed imaging and tremor disturbance, and finally obtain continuous, complete and distortion-free high-quality images. In order to obtain high-quality ultra-large wide format optical satellite images without distortion and seamless stitching, the construction and correction method of virtual imaging model for rotating area array mode is studied, and key technologies such as multi-gear virtual large area array CCD design and construction and virtual correction method based on ideal composite motion model are broken through to achieve distortion-free, seamless stitching and high-quality restoration compensation, providing comprehensive, accurate and rapid data and information support for subsequent on-orbit remote sensing applications.

Diffusing light modulated by focused ultrasound (FUS) or acousto optic (AO) sensing is a hybrid technique that utilizes ultrasound features, such as, deep penetration, localize and precise spatial resolution, to surpass optical scattering limitations in optical sensing and maintain the original optical contrast. In this paper, we compare the acousto-optic (AO) signal amplitude when using three different light sources; long coherence, laser diode (LD), and light emitting diode (LED). In recent years, there has been a growing trend towards the use of LEDs and laser diode LDs in optical applications because of their compact and small size, ease of use, safety features, and cost-effectiveness. Aim of this paper is to examine the capability of LDs and LEDs to be used in AO sensing. We evaluated differences in the tagging efficiency using detection of AO signal amplitude metrics. The results showed that particularly LDs are also capable of providing acceptable tagging efficiency in AO-based sensing compared with long coherent lasers and can be beneficial option for use in AO based techniques.

High-speed imaging capability is essential for photoacoustic computed tomography (PACT) to monitor biological dynamics. The filtered back-projection (FBP) algorithm is widely-used in PACT for image reconstruction owing to its high computational efficiency. To achieve high-speed imaging, several acceleration strategies based on the graphics processing units (GPUs) have been proposed to further increase the computational efficiency of the FBP algorithm. However, there are few acceleration strategies reported based on the multi-core central processing units (CPUs). Considering the fact that multi-core CPUs are much more accessible than high-performance GPUs, here we report a multi-core CPU-based framework for enhancing the computational efficiency of the FBP algorithm. In this framework, the highly-parallel back-projection part of the FBP algorithm is programed with C++ and implemented in parallel with multi-core CPUs. In addition, the pre-calculation strategy is applied in this framework to avoid unnecessary repetitive computations. The results show that implementing the back-projection part in parallel with C++ can reduce the image reconstruction time by a factor of 2.4 compared with the conventional implementation in which the FBP algorithm is fully programed with MATLAB and executed in parallel with parfor. By applying the pre-calculation, the image reconstruction time is further reduced by a factor of 2.2. Overall, the proposed framework increases the computational efficiency of the FBP algorithm by a factor of 5.5 and only takes 0.04 seconds to reconstruct an image with 512 × 512 pixels. This work is expected to promote the development of high-performance PACT systems that feature high imaging speed without GPUs.

Currently, as a new type of imaging device, the coated snapshot camera has advantages such as low cost, small size, high precision, and high frame rate. It has been widely used in industrial inspection, medical diagnosis, remote sensing, and other fields. In this paper, a detector control system based on Field-Programmable Gate Array (FPGA) is designed for the coated snapshot multispectral camera. In order to achieve high resolution and high frame rate, the detector control system adopts CMOS sensor GSENSE400BSI, which has a pixel count of 2048×2048, a pixel size of 11μm×11μm, and a frame rate of 46fps. This CMOS sensor has high integration, which is beneficial for simplifying the design of peripheral circuits for the detector. The hardware platform adopts a hierarchical structure with an image detector board, an FPGA main control board, and a power interface board. The FPGA chip serves as the logical control core of the detector control system, mainly completing high-speed data serialization and deserialization conversion, data buffering and processing, work mode switching, and interface logical control operations. In addition, the detector control system uses DDR3 for high-speed image data buffering, RS422 interface for serial communication, and Cameralink interface for outputting image data. Finally, the imaging results show that the image resolution of the detector control system reaches 2048×2048, and the frame rate remains stable at 25fps, achieving a good balance between imaging quality and speed. Subsequent work includes conducting actual tests and imaging experiments after coating the detector.

Structured illumination microscopy (SIM) stands out among full-field super-resolution imaging modes in life sciences because of its high imaging speed, low phototoxicity, and low photobleaching. Traditional SIM technology requires accurate illumination parameters of 9 original images to achieve artifact-free super-resolution image reconstruction. Currently, the most popular algorithm with excellent parameter estimation performance is the two-dimensional cross-correlation algorithm, which is implemented by a large number of cross-correlation calculations in each direction. However, this computationally intensive algorithm isn’t a better choice for the technical application of real-time and long-term live cell imaging. In this work, on the premise of ensuring the accuracy of parameter estimation and noise resistance, we propose a bisection-based parameter estimation algorithm that can reduce the number of cross-correlation calculations in each direction by an order of magnitude. In the algorithm, the whole pixel position of the wave vector is first determined. Then the cross-correlation value at both ends of the XY direction is calculated, and the larger cross-correlation value position and the middle position are taken as the position for the next cross-correlation value calculation, so as to gradually approach the actual wave vector position from coarse to fine. To verify the proposed algorithm, super-resolution image reconstruction for fluorescent samples was performed. The experimental results show that compared with traditional SIM algorithms, the proposed parameter estimation algorithm is more accurate and anti-noise, and less computationally intensive (with only about 1/10 of the original cross-correlation value), which is highly significant for the technical application of real-time and long-term live cell imaging.

The key technology of modern intelligent production is the concept of human-robot interaction. Operators and robots collaborate to perform complex tasks in various scenarios in such a system. Understanding human behavior introduces challenges for collaborative robot systems to perform efficient tasks in unstructured and dynamic environments. The proposed algorithm is based on classifying multi-class human motion from multimodal visual data. For this purpose, we apply the Laplace pyramid for visible, depth, and thermal data from imaging sensors. To prevent the appearance of visual artifacts, a multiscale approach is proposed for combining images based on the Laplace pyramid and calculating the optimal weights. The next stage is fused data preprocessing by the two-sided Gabor quaternionic Fourier transform and calculating 3D local binary dense micro-block difference. Also, we obtain human skeleton data using a convolutional neural network. Based on the coordinates of the unique points of the human body skeleton and the distances between them, a descriptor is constructed that describes the person's posture on each frame and considers the time information between frames. At the next stage, the 3D local binary dense micro-block difference and skeletal descriptor are combined into a single feature vector. Moreover, finally, this descriptor is classified by a neural network. We present simulation modeling on the effectiveness of the proposed action recognition algorithm in the RoboGuid environment.

Laser 3D imaging system has a very wide range of applications, such as satellite remote sensing and airborne surveillance. With the development of single photon detection technology, 3D imaging system based on single photon detection can meet these demanding requirements. In particular, single-photon detectors can provide single-photon sensitivity and ultra-high temporal resolution, and this high sensitivity allows the use of lower power lasers with longer detection distances. In this paper we present an active imaging system based on the single-photon ToF approach to obtain depth profiles of targets. The bistatic system comprised a pulsed laser source with an operational wavelength of 1064nm and an InGaAs/InP SPAD detector array which is highly efficient in the SWIR region.

Independent Component Analysis (ICA) applied to the image processing is the analysis of the characteristics of the human visual system based on sparse code. ICA provides a novel transfer domain method along multiple directions and offers excellent characteristic expression and an edge modeling feature. Color transfer is currently the best way to obtain natural colorization for grayscale image. By combining these two methods, our study focused on the exploration of an approach to natural color fusion by highlighting the corresponding band characteristics. A training image database was established and the characteristics of independent band were extracted to construct an analysis kernel and a synthesis kernel of the ICA domain. In the ICA domain, we applied regional energy fusion rule to generate a grayscale fusion image. According to the visual task, the source image was linearly projected to the color channels in brightness-color separation color space, giving color information to the grayscale fused image. Using a steerable pyramid, the sub-band image of source color image and the reference color image were generated and transferred with the mean and variance independently. Finally, a fused image with a daytime color appearance was obtained. The perception of the human eye and the objective evaluation indicated that the fused image highlights the band characteristics and enhances details with natural and visually pleasing colors, which further improves scene perception.

Non-obstructive azoospermia (NOA) is a severe male infertility condition characterized by impaired or absent sperm production in the testes. The primary treatment for NOA is microsurgical testicular sperm extraction (micro-TESE), which relies on accurately identifying healthy seminiferous tubules. In addressing this clinical need, we propose the utilization of mid-infrared photothermal (MIP) microscopy to identify spectroscopic signatures associated with NOA. Our preliminary results revealed that NOA tissues exhibited distinctive lipid distribution and reduced lipid peak intensity compared to tissues with normal sperm production. Leveraging principal components analysis (PCA), we successfully extracted key infrared spectroscopic features. When combined with logistic regression (LR), this approach achieved an impressive prediction accuracy of 95.0% in classifying testicular tissues. These findings highlight the potential of MIP microscopy in facilitating sperm retrieval by distinguishing seminiferous tubules based on their molecular composition.

Circular dichroism (CD) and asymmetric transmission (AT) are two outstanding chiroptical effects, which can be produced by planar chiral devices due to their distinct response to oppositely handed input light. The chiroptical effects have so far been used in realization of different applications including chiral imaging and sensing. This research realizes CD in the optical regime to attain AT for an optical wavelength regime by employing a supercell of hydrogenated amorphous Silicon nano-bars. The supercell is made of four achiral nanofins which are engineered in such a way that they exhibit a notable CD and AT. The design successfully achieves a CD of 55% and that of AT of 58% at the operating wavelength of 633 nm. The supercell absorbs the right circularly polarized light and transmits left circularly polarized light, thus behaving as a perfect circular polarization wave isolator i.e. the presented design can completely eradicate backscattering by minimizing reflectance.