This PDF file contains the front matter associated with SPIE Proceedings Volume 12523, including the Title Page, Copyright information, Table of Contents, and Conference Committee information.

Research supporting improved anomaly detection performance benefits a wide range of technical applications, and thus, the definition of what anomalies are and the subsequent means to detect them are wide ranging. In this treatment, an overview of the development of an anomaly detection approach based on spectral signatures obtained with hyperspectral unmixing is presented. The algorithm is designed to address some of the shortcomings of current techniques whose functionality is dependent upon normalized differences between discrete frequencies or spectral components, or those based on estimated distances between background spectra and pixels under test. Details about the extracted endmembers and their use for effective anomaly detection will be presented as well as, some thoughts on the expected requirements for future machine learning based implementations.

Interaction of light with biological material constitutes a powerful tool for example in medical diagnosis. For example, the degree of linear polarization (DoLP) allows to differentiate between healthy and cancerous tissue. Polarimetric imaging by means of space-division techniques has become recently popular due to advances in nanotechnology and commercial availability of microgrid polarizers integrated onto the sensing chip of a camera. In order to obtain the linear Stokes parameters from a sample in a single shot, we developed a microscopy setup incorporating a polarized sensor in the imaging plane. Besides, by means of illumination multiplexing with linear States of Polarization in the input, Mueller matrix elements can be retrieved in a single shot. Experimental results for measures of DoLP and AoP as well as Mueller matrix elements over tissue samples are presented.

The design of a one-dimensional lens is presented in this article. The method used for this design is based on geometric optics, where refracting surfaces are used to redirect rays of light. It is known that the illumination profile of many LEDs can approximately be modeled by Lambertian distribution, mainly depending on the angle of the emitted ray with the normal to the surface. We want to construct a lens which is to be placed at a reasonable distance away from an LED source, in such a way that the radiance profile of this source is refracted at both surfaces of the lens so that the illumination profile on the target surface is as uniform as possible. This task requires the optimization of a certain error function. The main constraint of the problem is to preserve the prescribed input and output intensities so that the requirement of total energy preservation is fulfilled.

Multifocus acquisition provides a way to reconstruct high resolution images from a 3D scene by capturing a sequence of multiple focal planes along the optical axis of the camera. Since the capture of multifocus images is not simultaneous but relies on a temporal sequence of captures while focusing at different focal planes the multifocus method is usually limited to static scenes. However since the development of high speed electrically focus tunable lenses it is also possible to reconstruct dynamic scenes from multifocus acquisition provided that the time to perform a complete sweep along z axis is much shorter than the time involved in the dynamic scene. In the present work we propose all-in-focus reconstruction of a 3D dynamic scene from multiple-multifocus captures.

Digital adaptive optics created using homodyne encoding can mitigate atmospheric turbulence in passive imaging systems. This work demonstrates a self-referencing homodyne interferometry technique that combines the passive imaging utility of multi-frame algorithmic procedures with the single-frame correction capability of the Shack-Hartmann adaptive optics technique. As an expansion of recent progress on three sub-aperture assemblies, this work showcases the latest results from the NIWC Pacific team to include coherently reconstructing 18 sub-apertures.

Over the past decades growing attention has been paid to the effect of propagation of waves in different types of randomly inhomogeneous media. This increasing interest is due, among other factors, to the novel methods that are being proposed for imaging small objects in cluttered environments and the need for good solvers which can correctly model fluctuations effects in these media. This is a major challenge in many research areas such as medical imaging, remote sensing, nondestructive testing and wireless communications, etc. In these applications, distinctly different mathematical models are used for imaging in continuous and discrete random media. The discrete random media are used to describe multiple scattering in continuous random media but it is unlikely that discrete inclusions are always adequate to represent inhomogeneities that have no clear boundaries between their different components. Imaging becomes very difficult to perform in discrete random media when resonance with multiple scattering exists. Resonance causes image distortion arising from the underlying interactions of multiply scattered waves at resonance frequencies. This article presents a numerical study of wave propagation in the existence of resonance with multiple scattering using the Foldy-Lax-Lippmann-Schwinger formalism, which was employed for the multiply scattered waves. We use a two-body simulation to demonstrate the imaging problem caused by resonance with multiple scattering and to understand why it can hardly be simulated with the original Foldy-Lax model. We demonstrate the result of removing the sharp responses in the resonance regime to recover the damaged images.

Digital holographic microscopy has been widely adopted in the field of biology because of its ability of quantifying phase information. Unlike those conventional transmission-type phase imaging techniques, reflective digital holographic microscopy can be applied in the field of display industry or semiconductor industry for inspecting unwanted defects in a certain fabrication process. Consequently, it is necessary and required to achieve large field of view and resolution enhancement to meet the demands on the fast and accurate shape measurement. To increase the minimum resolution, we are going to adopt programmable illumination patterns by the use of fast-activating digital micro-mirror devices (DMDs) referring to Fourier ptychographic illumination. In this work, to expand space bandwidth product in reflection-type digital holographic microscopy (RDHM) configuration, we apply a fast-activating digital micro-mirror device for a amplitude-type spatial light modulator and a micro-lens array in the reference arm of a conventional RDHM.

Convolutional neural networks (CNNs) are a widely researched neural network architecture that has demonstrated exemplary performance in image processing tasks and applications compared to other popular deep learning and machine learning methods resulting in state-of-the-art performance in many image processing tasks such as image classification and segmentation. CNNs operate on the principle of automated learning of filters or kernels in contrast with hand-crafted digital filters to extrapolate features from images effectively. This paper aims to investigate whether a matrix's determinant can be used to preserve information in CNN convolutional layers. Geometrically the absolute value of the determinant is defined as a scaling factor of the linear transformation resulting from matrix multiplication. When an image's size is reduced into a feature space through a convolutional layer of a CNN, some information is lost. The intuition is that the scaling factor that results from the determinant of the pooling layer matrix can enhance the feature space introducing scaling as a piece of information in the feature space as well as lost relations between adjacent pixels.

In medical and microscopy imaging applications where the object is not directly visible, images are never identical to the ground truth. In three-dimensional structured illumination microscopy (3D-SIM), acquired images taken from the object have limited resolution due to the the point spread function (PSF) of the imaging system. Additionally, due to the data acquisition process, images taken under low light and in the presence of electrooptical noise can have a low signal-to-noise ratio as well as suffer from other undesirable aberrations. To obtain a high-resolution restored image, the data must be digitally processed. The inverse imaging problem in 3D-SIM has been solved using various computational imaging techniques. Traditional model-based computational approaches can result in image artifacts due to required, yet not accurately known system parameters. Furthermore, some iterative computational imaging methods can be computationally intensive. Deep learning (DL) approaches, as opposed to traditional image restoration methods, can tackle the issue without access to the analytical model. Although some are effective, they are biased since they do not use the 3D-SIM model. This research aims to provide an unrolled physics-informed (UPI) generative adversarial network (UPIGAN) for reconstructing 3D-SIM images utilizing data samples of mitochondria from a 3D-SIM system. This design uses the benefits of physics knowledge in the unrolling step. Moreover, the GAN employs a Residual Channel Attention super-resolution deep neural network (DNN) in its generator architecture. The results from both a qualitative and quantitative comparison, present a positive impact on the reconstruction when the UPI term is used in the GAN versus using the GAN architecture without it.

Small angle x-ray scatter from microstructure too small to resolve in conventional imaging can provide an additional “dark field” signal that is complementary to attenuation and phase. Unfortunately, the low spatial coherence of clinical sources reduces dark field contrast. Focusing polycapillary optics are employed to allow for the use of high-power primary sources by increasing the phase signal after the focus. The method for this system is to structure the beam with a low-cost wire mesh that further relaxes the coherence requirement on the source. However, the coarseness of the mesh reduces the strength of the dark field signal compared with grating-based techniques, so optimization of the signal is important.

X-ray phase and dark field imaging offer two additional channels of information to enhance image contrast as well as providing information on material micro-texture that is unavailable to conventional x-ray imaging. These signals are commonly acquired by using multiple precisely aligned, fine-pitched gratings to both pattern the beam and to detect subtle shifts and blurring in this pattern. We instead use a single, low-cost, easily-aligned, coarse-pitched mesh to produce a pattern that is imaged directly to produce phase and dark field computed tomography (CT) images. We demonstrate phase and CT reconstructions using our system for a variety of phantoms.

For many imaging applications installed on moving platforms, especially for smaller platforms without the advantage of large inertial mass, jitter is one the primary drivers in modulation transfer function (MTF) degradation. The use of an inertial measurement unit (IMU) to detect motion for non-blind deconvolution of imagery is not a new concept. However, most systems are focused on still photographs for small optical systems, such as cell phones, and are not focused on real-time implementation for full motion video (FMV) with dynamic systems that have multiple moving parts. Further, no existing system utilizes IMU information to intelligently decide when a camera system should integrate during the typical 33 ms framerate. Having control over when to integrate and for how long allows for greater signal-to-noise ratios (SNRs) at smaller jitters by selecting times when jitter is minimized based on the LOS motion. This work presents the framework of a jitter mitigation approach that both optimally decides on the integration window and implements non-blind deconvolution to produce a system that provides enhanced image resolution under a variety of conditions

Photoacoustic microscopy (PAM) is a non-invasive, label-free functional imaging technique that provides high absorption contrast with high spatial resolution. Spatial sampling density and data size are key determinants of PAM imaging speed. Therefore, under sampling methods that reduce the number of scan points are usually employed to improve the imaging speed of PAM by increasing the scan step size. Because under sampling techniques sacrifice spatial sampling density, deep learning-based reconstruction techniques have been explored as alternatives. However, these methods have been applied to reconstruct two-dimensional PAM images related to spatial sampling density. Therefore, by considering the number of data points, the data size, and the characteristics of PAM to provide three-dimensional (3D) volume data, this study proposes a deep-learning-based complete reconstruction of under sampled 3D PAM data. newly reported to Obtained from real experiments (i.e. not manually generated). Quantitative analysis results show that the proposed method exhibits robustness and outperforms interpolation-based reconstruction methods at various under sampling ratios, resulting in 80x faster imaging speed and 800x smaller data. Improves PAM system performance with size. Furthermore, the applicability of this method is experimentally verified by enlarging a sparsely sampled test dataset. His proposed deep learning-based PAM data reconstruction has been demonstrated to be the closest model available under experimental conditions, significantly reducing the data size for processing and effectively reducing the imaging time.

For reconstruction in spatial compressive imaging, we use a module to fuse and extract the information in sampling templates, this obtained feature vector becomes the attention weight, which is multiplied with the feature maps of the compressed measurement frames. In addition, unlike previous networks using segmented images, we use full measurement frames collected as our network input. Thus the local information of objects can be preserved and blocky effect can be avoid. We have tested the network performance on the datasets, Set5, Set14, BSD100, Urban100, Manga109, with 25% compression rate, respectively. We obtain the PSNR\SIM values in the range, [26.5dB, 31.9dB]\[0.82, 0.90]. This result is better than [23.6dB, 29.0dB]\[0.72, 0.85] obtained using the best algorithms in the same application based on our knowledge.

We proposed a variable circular coded aperture super-resolution imaging method to tackle the imaging limitation under the Nyquist sampling frequency insufficiency. By adjusting the circular aperture size to modulate the point spread function (PSF), a sequence of images captured by a low-resolution imaging detector. High-resolution image reconstructed by the iterative algorithm from a serial of captured low-resolution images. The simulation results show that our method can improve the resolution by 1.78 times. Without changing the imaging system or increasing its complexity, we adopt wavefront encoding by adding a circular aperture in front of the imaging lens to achieve super-resolution imaging. This imaging method is expected to apply the coded-aperture super-resolution imaging to miniaturized devices to upgrade it imaging quality.

Far-field photoelectric detection has long been powerful tool for defense and public security. The resolution, one of the most critical metrics of a detection system, characterizes its capability to discern details and is primarily limited by the aperture dimension. In addition to expanding the aperture directly, aperture synthesis is a typical method for resolution enhancement, in which a series of sub-apertures form a larger virtual aperture. Here we show a super-resolution imaging method based on aperture synthesis with incoherent illumination. The phase of the light field is retrieved with an autocorrelator, and details, which could not be captured with the same aperture diameter, can be obtained under different sub-apertures by wavefront modulation. The super-resolution result is reconstructed by synthesizing the sub-aperture images in the Fourier domain using an iterative algorithm. The capability in resolution enhancement of this method is demonstrated with simulations and experiments.