Recovering badly damaged face images is a useful yet challenging task, especially in extreme cases where the masked or damaged region is very large. One of the major challenges is the ability of the system to generalize on faces outside the training dataset. We propose to tackle this extreme inpainting task with a conditional Generative Adversarial Network (GAN) that utilizes structural information, such as edges, as a prior condition. Edge information can be obtained from the partially masked image and a structurally similar image or a hand drawing. In our proposed conditional GAN, we pass the conditional input in every layer of the encoder while maintaining consistency in the distributions between the learned weights and the incoming conditional input. We demonstrate the effectiveness of our method with badly damaged face examples.

SARS-CoV-2 is a highly contagious, airborne-transmission, virus that can be spread by people who do not have obvious symptoms. In 2020, that combination of features forced much of the world to impose a wide variety of forms of social distancing, ranging from simple recommendations restricting how shared spaces can be used to rigidly enforced quarantines. It is unclear how much distancing is enough, but it is clear that the economic and emotional costs of distancing are high. Fortunately, consistent use of simple face masks dramatically reduces the probability of others becoming infected. The catch is that a significant fraction of the US population either is refusing to wear masks or is wearing masks in ways that render them ineffective. For example, it is problematic for a shop owner to prevent potential customers who are not properly masked from entering their store. Thus, we have created the Covered Safe Entry Scanner–an open source system that uses image processing methods to automatically check for proper use of masks and potentially deny entry to those who do not comply. This paper describes the design, algorithms, and performance of the mask recognition system.

In most optic systems images are captured using a CCD/CMOS sensor, where the phases of the converted photons are inevitably lost. Fourier Ptychographic Microscopy (FPM) circumvents this issue by capturing normal microscopy images, and Fourier transforming them computationally (hence the name). Reconstructing the complex object not only yields amplitude but also phase information, enhanced up to super-resolution. Yet one disadvantage remains unsolved: FPM is a very ill-posed problem, the algorithm is not guaranteed to converge to the correct solution, if it converges at all. In practice this means that there is reasonable doubt if the recovered image actually represents the object under the microscope. This work inquires the quality of FPM reconstruction under variation of important system parameters in simulation and experiment. It shows that the alignment of the illumination source is quite critical: even 0.2 degrees off renders reconstruction useless. This paper thus furthers the costbenefit analysis of which amount of computation time should be spent on digital post-correction.

We propose a neural network architecture combined with specific training and inference procedures for linear inverse problems arising in computational imaging to reconstruct the underlying image and to represent the uncertainty about the reconstruction. The proposed architecture is built from the model-based reconstruction perspective, which enforces data consistency and eliminates the artifacts in an alternating manner. The training and the inference procedures are based on performing approximate Bayesian analysis on the weights of the proposed network using a variational inference method. The proposed architecture with the associated inference procedure is capable of characterizing uncertainty while performing reconstruction with a modelbased approach. We tested the proposed method on a simulated magnetic resonance imaging experiment. We showed that the proposed method achieved an adequate reconstruction capability and provided reliable uncertainty estimates in the sense that the regions having high uncertainty provided by the proposed method are likely to be the regions where reconstruction errors occur.

Conventional X-ray computed tomography (CT) systems obtain single- or dual-energy measurements, from which dual-energy CT has emerged as the superior way to recognize materials. Recently photon counting detectors have facilitated multi-spectral CT which captures spectral information by counting photon arrivals at different energy windows. However, the narrow energy bins result in a lower signal-to-noise ratio in each bin, particularly in the lower energy bins. This effect is significant and challenging when high-attenuation materials such as metal are present in the area to be imaged. In this paper, we propose a novel technique to estimate material properties with multi-spectral CT in the presence of high-attenuation materials. Our approach combines basis decomposition concepts using multiple-spectral bin information, as well as individual energy bin reconstructions. We show that this approach is robust in the presence of metal and outperforms alternative techniques for material estimation with multi-spectral CT as well with the state-of-art dual-energy CT.

A Supervised Learning Approach for Dynamic Sampling (SLADS) addresses traditional issues with the incorporation o stochastic processes into a compressed sensing method. Statistical features, extracted from a sample reconstruction, estimate entropy reduction with regression models, in order to dynamically determine optimal sampling locations. This work introduces an enhanced SLADS method, in the form of a Deep Learning Approach for Dynamic Sampling (DLADS), showing reductions in sample acquisition times for high-fidelity reconstructions between ˜ 70–80% over traditional rectilinear scanning. These improvements are demonstrated for dimensionally asymmetric, high-resolution molecular images of mouse uterine and kidney tissues, as obtained using Nanospray Desorption Electro- Spray Ionization (nano-DESI) Mass Spectrometry Imaging (MSI). The methodology for training set creation is adjusted to mitigate stretching artifacts generated when using prior SLADS approaches. Transitioning to DLADS removes the need for feature extraction, further advanced with the employment of convolutional layers to leverage inter-pixel spatial relationships. Additionally, DLADS demonstrates effective generalization, despite dissimilar training and testing data. Overall, DLADS is shown to maximize potential experimental throughput for nano-DESI MSI.

Tunable diode laser absorption tomography (TDLAT) has emerged as a popular nonintrusive technique for simultaneous sensing of gas concentration and temperature by making light absorbance measurements. Major challenge of TDLAT imaging is that the measurement data is very sparse. Therefore, precise models are required to describe the measurement process (forward model) and the behavior of the gas flow properites (prior model) to get accurate reconstructions. The sparsity of the measurement data makes TDLAT very sensitive to the accuracy of the models and makes it prone to overfitting. Both the forward and prior models can have systematic errors due to several reasons. So far, substantial amount of work has been done by researchers on developing reconstruction methods and formulating models, forward and prior. Yet, there has not been significant research work done on constructing a metric for goodness of the model fit that can indicate when there is an inaccuracy in the forward or the prior model. In this paper, we present a metric for goodness of model fit that can be used to indicate if the models used in the reconstruction are inaccurate. Results show that our metric can reliably quantify the goodness of model fit for sparese data reconstruction problems such as TDLAT.

Dual-energy computed tomography (DECT) has been widely used to reconstruct basis components. In previous studies, ou DECT algorithm has shown high accuracy in stopping power ratio (SPR) estimation of fixed objects for proton radiotherapy planning. However, patient movement between sequential data acquisitions may lead to severe motion artifacts in the component images. In order to reduce or eliminate the motion artifacts in clinical applications, we combine a deformable registration method with an accurate joint statistical iterative reconstruction algorithm, dual-energy alternating minimization (DEAM). Image registration is a process of geometrically aligning two or more images. We implement a multi-modality symmetric deformable registration method based on Advanced Normalization Tools (ANTs) to automatically align the scans we acquire for the same patient. The precalculated registration mapping and its inverse are then embedded into each iteration of the DEAM algorithm. The performance of warped DEAM is quantitatively assessed. Theoretically, the performance of warped DEAM on moved patients should be comparable to the performance of the original DEAM algorithm on fixed objects. The warped DEAM algorithm reduces motion artifacts while preserving the accuracy of the iterative joint statistical CT reconstruction algorithm, which enables us to reconstruct accurate results from sequentially scanned dual-energy patient data.

The performance of a convolutional neural network (CNN) on an image texture detection task as a function of linear image processing and the number of training images is investigated. Performance is quantified by the area under (AUC) the receiver operating characteristic (ROC) curve. The Ideal Observer (IO) maximizes AUC but depends on high-dimensional image likelihoods. In many cases, the CNN performance can approximate the IO performance. This work demonstrates counterexamples where a full-rank linear transform degrades the CNN performancebelow the IO in the limit of large quantities of training dataand network layers. A subsequent linear transform changes theimages’ correlation structure, improves the AUC, and again demonstrates the CNN dependence on linear processing. Compression strictly decreases or maintains the IO detection performance while compression can increase the CNN performance especially for small quantities of training data. Results indicate an optimal compression ratio for the CNN based on task difficulty, compression method, and number of training images. c 2020 Society for Imaging Science and Technology.