Spatially varying defocus map estimation from a single image based on spatial aliasing sampling method.

In current optical systems, defocus blur is inevitable due to the constrained depth of field. However, it is difficult to accurately identify the defocus amount at each pixel position as the point spread function changes spatially. In this paper, we introduce a histogram-invariant spatial aliasing sampling method for reconstructing all-in-focus images, which addresses the challenge of insufficient pixel-level annotated samples, and subsequently introduces a high-resolution network for estimating spatially varying defocus maps from a single image. The accuracy of the proposed method is evaluated on various synthetic and real data. The experimental results demonstrate that our proposed model outperforms state-of-the-art methods for defocus map estimation significantly.

MTF improvement for optical synthetic aperture system via mid-frequency compensation.

Optical synthetic aperture imaging system has grown out the quest for higher angular resolution in astronomy, which combines the radiation from several small sub-apertures to obtain a resolution equivalent to that of a single filled aperture. Due to the discrete distribution of the sub-apertures, pupil function is no longer a connected domain, which further leads to the attenuation or loss of the mid-frequency modulation transfer function (MTF). The mid-frequency MTF compensation is therefore a key focus. In this paper, a complete mid-frequency compensation algorithm is proposed, which can extract and fuse the frequency of different synthetic aperture systems and monolithic aperture systems according to their special MTF characteristics. The dimensions of the monolithic aperture and optical synthetic aperture system are derived, and the longest baseline of the monolithic aperture is much smaller than that of the optical synthetic aperture system. Then the separated spatial frequency information is extracted and synthesized according to the spatial frequency equivalence point. Finally, the full-frequency enhanced image is recovered by using improved Wiener-Helstrom filter, which adopts specific parameters based on different sub-aperture arrangements. The mid-frequency MTF of Golay-3 increases from 0.12 to 0.16 and that of Golay-6 increases from 0.06 to 0.18. Both the simulation and experiment prove that the proposed method not only realizes the spatial resolution determined by the longest baseline of the optical synthetic aperture system, but also successfully compensates its mid-frequency MTF.

Optimized Golay-9 array configurations for mid-frequency compensation in optical sparse aperture systems.

Optical sparse aperture (OSA) imaging systems show great potential to generate higher resolution images than those of equivalent single filled aperture systems. However, due to the sparsity and dispersion of sparse aperture arrays, pupil function is no longer a connected domain, which further attenuates or loses the mid-frequency modulation transfer function (MTF), resulting in lower mid-frequency contrast and blurred images. Therefore, an improved traversal algorithm is proposed to optimize Golay-9 array configurations for compensating the mid-frequency MTF. Its structural parameters include diameters of sub-apertures, relative rotation angles between individual sub-apertures, and radius of concentric circles. Then, these parameters are traversed successively in order. Finally, the influences of the obtained optimized array configurations on the mid-frequency MTF are analyzed in detail, and the image performances are evaluated. The experimental results prove the contrast enhancement. Compared with a Golay-9 array at F=36.5%, the maximum MTF increases from 0.1503 to 0.307, and the mid-frequency MTF is boosted from 0.0565 to 0.0767. In addition, the peak signal to noise ratio of the degraded image is promoted from 19.75 dB to 20.63 dB. Both quantitative and qualitative evaluations demonstrate the validity of the proposed method.

Image restoration for synthetic aperture systems with a non-blind deconvolution algorithm via a deep convolutional neural network.

Optical synthetic aperture imaging systems, which consist of in-phase circular sub-mirrors, can greatly improve the spatial resolution of a space telescope. Due to the sub-mirrors' dispersion and sparsity, the modulation transfer function is decreased significantly compared to a fully filled aperture system, which causes obvious blurring and loss of contrast in the collected image. Image restoration is the key to get the ideal clear image. In this paper, an appropriative non-blind deconvolution algorithm for image restoration of optical synthetic aperture systems is proposed. A synthetic aperture convolutional neural network (CNN) is trained as a denoiser prior to restoring the image. By improving the half-quadratic splitting algorithm, the image restoration process is divided into two subproblems: deconvolution and denoising. The CNN is able to remove noise in the gradient domain and the learned gradients are then used to guide the image deconvolution step. Compared with several conventional algorithms, scores of evaluation indexes of the proposed method are the highest. When the signal to noise ratio is 40 dB, the average peak signal to noise ratio is raised from 23.7 dB of the degraded images to 30.8 dB of the restored images. The structural similarity index of the results is increased from 0.78 to 0.93. Both quantitative and qualitative evaluations demonstrate that the proposed method is effective.

Object-independent piston diagnosing approach for segmented optical mirrors via deep convolutional neural network.

Piston diagnosing approaches based on neural networks have shown great success, while a few methods are heavily dependent on the imaging target of the optical system. In addition, they are inevitably faced with the interference of submirrors. Therefore, a unique object-independent feature image is used to form an original kind of data set. Besides, an extremely deep image-based convolutional neural network (CNN) of 18 layers is constructed. Furthermore, 9600 images are generated as a data set for each submirror with a special measure of sensitive area extracting. The diversity of results among all the submirrors is also analyzed to ensure generalization ability. Finally, the average root mean square error of six submirrors between the real piston values and the predicted values is approximately 0.0622λ. Our approach has the following characteristics: (1) the data sets are object-independent and contain more effective details, which behave comparatively better in CNN training; (2) the complex network is deep enough and only a limited number of images are required; (3) the method can be applied to the piston diagnosing of segmented mirror to overcome the difficulty brought by the interference of submirrors. Our method does not require special hardware, and is fast to be used at any time, which may be widely applied in piston diagnosing of segmented mirrors.

Breadth-first piston diagnosing approach for segmented mirrors through supervised learning of multiple-wavelength images.

Piston diagnosing approaches for segmented mirrors via machine-learning have shown great success. However, they are inevitably challenged with 2π ambiguity, and the accuracy is usually influenced by the location and number of submirrors. A piston diagnosing approach for segmented mirrors, which employs the breadth-first search (BFS) algorithm and supervised learning strategies of multi-wavelength images, is investigated. An original kind of object-independent and normalized dataset is generated by the in-focal and defocused images at different wavelengths. Additionally, the segmented mirrors are divided into several sub-models of binary tree and are traversed through the BFS algorithm. Furthermore, two deep image-based convolutional neural networks are constructed for predicting the ranges and values of piston aberrations. Finally, simulations are performed, and the accuracy is independent of the location and number of submirrors. The Pearson correlation coefficients for test sets are above 0.99, and the average root mean square error of segmented mirrors is approximately 0.01λ. This technique allows the piston error between segmented mirrors to be measured without 2π ambiguity. Moreover, it can be used for data collected by a real setup. Furthermore, it can be applied to segmented mirrors with different numbers of submirrors based on the sub-model of a binary tree.

Edge detection for optical synthetic apertures based on conditional generative adversarial networks.

Detecting the interference fringes of the optical synthetic aperture is the core in preventing misalignments of the sub-mirrors in piston, tip, and tilt. These fringes are characterized as follows: (1) the edge information of sub-mirrors is accompanied by complex shapes and large gaps; and (2) the traditional edge detection algorithms have different optimal thresholds under different interference fringes, and they may lose boundary information. To address these problems, a novel method for detecting the edge of synthetic aperture fringe images is proposed. Because conditional generative adversarial networks avoid the difficulty of designing the loss function for specific tasks, they are suitable for our project. We trained over 8000 images based on real images and simulated images. Experiments prove that the proposed method can reduce the false detection rate to 0.2, compared with 0.56 by Canny algorithm. This method can also directly detect the fringe edge of the optical synthetic aperture systems, which are accompanied by varied shapes and a growing number of sub-mirrors. When the input images lose boundary information, the traditional algorithm does not restore the boundary, but the proposed method makes a decision globally, and thus it guesses and then fills the boundary. The maximum error of the generated boundary and the actual boundary is two pixels.

Increasing aperture and depth of field simultaneously with wavefront coding technology.

We present a method that simultaneously increases the aperture and depth of field (DOF). A novel, to the best of our knowledge, method called lens-combined wavefront coding (WFC) is proposed for optical design. By rationally balancing rather than minimizing the aberrations, the DOF enhances instead of reduces with expansion of the aperture size. Two optical systems by traditional design and lens-combined WFC are designed to demonstrate our concept. Experiments are conducted using the manufactured lenses to show the extension of the DOF both in the laboratory and the real scene. A spatially invariant deconvolution algorithm is exploited to further suppress the aberrations regarding the field of view. The results show that the aperture and the DOF can be successfully enhanced at the same time.

Rotating asymmetrical phase mask method for improving signal-to-noise ratio in wavefront coding systems.

We present a method that combines an asymmetrical phase mask (PM) and its rotated PM to potentially improve the signal-to-noise ratio in wavefront coding systems. The property of the rotated PM is analyzed. A complementary promotional synthetic optical transfer function is generated as the deconvolution filter for image recovery. The image artifacts are suppressed through processing in the frequency domain. Simulation results provide a quantitative analysis that is based on a cubic PM. Experimental results show that our method can improve image quality over a wide range of defocus.

Image quality improvement by the structured light illumination method in an optical readout cantilever array infrared imaging system.

The structured light illumination method is applied in an optical readout uncooled infrared imaging system to improve the IR image quality. The unavoidable nonuniform distribution of the initial bending angles of the bimaterial cantilever pixels in the focal plane array (FPA) can be well compensated by this method. An ordinary projector is used to generate structured lights of different intensity distribution. The projected light is divided into patches of rectangular regions, and the brightness of each region can be set automatically according to the deflection angles of the FPA and the light intensity focused on the imaging plane. By this method, the FPA image on the CCD plane can be much more uniform and the image quality of the IR target improved significantly. A comparative experiment is designed to verify the effectiveness. The theoretical analysis and experimental results show that the proposed structured light illumination method outperforms the conventional one, especially when it is difficult to perfect the FPA fabrication.

High-speed infrared imaging by an uncooled optomechanical focal plane array.

In this paper, we theoretically and experimentally demonstrate that the imaging speed of the optomechanical focal plane array infrared imaging system can be significantly improved by changing the pressure in the vacuum chamber. The decrease in the thermal time constant is attributed to the additional thermal conductance caused by air. The response time will be greatly shortened to about 1/3 time in low vacuum (around ∼10(2) Pa) compared with that in high vacuum. At a chamber pressure of 50 Pa, the "trailing" in the IR image of a moving hot iron is eliminated with negligible deterioration in the image quality. Moreover, infrared images on rapid occurrence events, such as ignition of an alcohol blast burner, lighting and fusion of a tungsten filament, are captured at a frame rate up to 200 Hz. The above results show that the proposed pressure-dependent performance provides a way to improve the system imaging speed and helps to slow down a dynamic event, which is of great value to the uncooled IR imaging systems in practical applications.

Imaging and image restoration of an on-axis three-mirror Cassegrain system with wavefront coding technology.

Wavefront coding (WFC) technology is adopted in the space optical system to resolve the problem of defocus caused by temperature difference or vibration of satellite motion. According to the theory of WFC, we calculate and optimize the phase mask parameter of the cubic phase mask plate, which is used in an on-axis three-mirror Cassegrain (TMC) telescope system. The simulation analysis and the experimental results indicate that the defocused modulation transfer function curves and the corresponding blurred images have a perfect consistency in the range of 10 times the depth of focus (DOF) of the original TMC system. After digital image processing by a Wiener filter, the spatial resolution of the restored images is up to 57.14 line pairs/mm. The results demonstrate that the WFC technology in the TMC system has superior performance in extending the DOF and less sensitivity to defocus, which has great value in resolving the problem of defocus in the space optical system.

Priors-assisted dehazing network with attention supervision and detail preservation.

Single image dehazing is a challenging computer vision task for other high-level applications, e.g., object detection, navigation, and positioning systems. Recently, most existing dehazing methods have followed a "black box" recovery paradigm that obtains the haze-free image from its corresponding hazy input by network learning. Unfortunately, these algorithms ignore the effective utilization of relevant image priors and non-uniform haze distribution problems, causing insufficient or excessive dehazing performance. In addition, they pay little attention to image detail preservation during the dehazing process, thus inevitably producing blurry results. To address the above problems, we propose a novel priors-assisted dehazing network (called PADNet), which fully explores relevant image priors from two new perspectives: attention supervision and detail preservation. For one thing, we leverage the dark channel prior to constrain the attention map generation that denotes the haze pixel position information, thereby better extracting non-uniform feature distributions from hazy images. For another, we find that the residual channel prior of the hazy images contains rich structural information, so it is natural to incorporate it into our dehazing architecture to preserve more structural detail information. Furthermore, since the attention map and dehazed image are simultaneously predicted during the convergence of our model, a self-paced semi-curriculum learning strategy is utilized to alleviate the learning ambiguity. Extensive quantitative and qualitative experiments on several benchmark datasets demonstrate that our PADNet can perform favorably against existing state-of-the-art methods. The code will be available at https://github.com/leandepk/PADNet.

A three-stage deep learning-based training frame for spectra baseline correction.

For spectrometers, baseline drift seriously affects the measurement and quantitative analysis of spectral data. Deep learning has recently emerged as a powerful method for baseline correction. However, the dependence on vast amounts of paired data and the difficulty in obtaining spectral data limit the performance and development of deep learning-based methods. Therefore, we solve these problems from the network architecture and training framework. For the network architecture, a Learned Feature Fusion (LFF) module is designed to improve the performance of U-net, and a three-stage training frame is proposed to train this network. Specifically, the LFF module is designed to adaptively integrate features from different scales, greatly improving the performance of U-net. For the training frame, stage 1 uses airPLS to ameliorate the problem of vast amounts of paired data, stage 2 uses synthetic spectra to further ease reliance on real spectra, and stage 3 uses contrastive learning to reduce the gap between synthesized and real spectra. The experiments show that the proposed method is a powerful tool for baseline correction and possesses potential for application in spectral quantitative analysis.

Non-contact detection of oxygen saturation based on visible light imaging device using ambient light.

A method that remotely measures blood oxygen saturation through two cameras under regular lighting is proposed and experimentally demonstrated. Two narrow-band filters with their visible wavelength of 660nm and 520nm are mounted to two cameras respectively, which are then used to capture two photoplethysmographic (PPG) from the subject simultaneously. The data gathered from this system, including both blood oxygen saturation and heart rate, is compared to the output of a traditional figure blood volume pulse (BVP) senor that was employed on the subject at the same time. Result of the comparison showed that the data from the new, non-contact system is consistent and comparable with the BVP senor. Compared to other camera-based measuring method, which requires additional close-up lighting, this new method is achievable under regular lighting condition, therefore more stable and easier to implement. This is the first demonstration of an accurate video-based method for non-contact oxygen saturation measurements by using ambient light with their respective visible wavelength of 660nm and 520nm which is free from interference of the light in other bands.

Prediction of Impulsive Aggression Based on Video Images.

In response to the subjectivity, low accuracy, and high concealment of existing attack behavior prediction methods, a video-based impulsive aggression prediction method that integrates physiological parameters and facial expression information is proposed. This method uses imaging equipment to capture video and facial expression information containing the subject's face and uses imaging photoplethysmography (IPPG) technology to obtain the subject's heart rate variability parameters. Meanwhile, the ResNet-34 expression recognition model was constructed to obtain the subject's facial expression information. Based on the random forest classification model, the physiological parameters and facial expression information obtained are used to predict individual impulsive aggression. Finally, an impulsive aggression induction experiment was designed to verify the method. The experimental results show that the accuracy of this method for predicting the presence or absence of impulsive aggression was 89.39%. This method proves the feasibility of applying physiological parameters and facial expression information to predict impulsive aggression. This article has important theoretical and practical value for exploring new impulsive aggression prediction methods. It also has significance in safety monitoring in special and public places such as prisons and rehabilitation centers.

Multi-Class Wound Classification via High and Low-Frequency Guidance Network.

Wound image classification is a crucial preprocessing step to many intelligent medical systems, e.g., online diagnosis and smart medical. Recently, Convolutional Neural Network (CNN) has been widely applied to the classification of wound images and obtained promising performance to some extent. Unfortunately, it is still challenging to classify multiple wound types due to the complexity and variety of wound images. Existing CNNs usually extract high- and low-frequency features at the same convolutional layer, which inevitably causes information loss and further affects the accuracy of classification. To this end, we propose a novel High and Low-frequency Guidance Network (HLG-Net) for multi-class wound classification. To be specific, HLG-Net contains two branches: High-Frequency Network (HF-Net) and Low-Frequency Network (LF-Net). We employ pre-trained models ResNet and Res2Net as the feature backbone of the HF-Net, which makes the network capture the high-frequency details and texture information of wound images. To extract much low-frequency information, we utilize a Multi-Stream Dilation Convolution Residual Block (MSDCRB) as the backbone of the LF-Net. Moreover, a fusion module is proposed to fully explore informative features at the end of these two separate feature extraction branches, and obtain the final classification result. Extensive experiments demonstrate that HLG-Net can achieve maximum accuracy of 98.00%, 92.11%, and 82.61% in two-class, three-class, and four-class wound image classifications, respectively, which outperforms the previous state-of-the-art methods.

A novel optical readout infrared FPA imaging system with fiber reference channel.

A novel fiber reference optical readout method was proposed in the bi-material micro cantilever infrared imaging system, which consists of an infrared imaging channel, an optical readout channel and a fiber reference channel. The fiber reference channel is used to monitor the intensity fluctuation of the light source, and provide a signal to correct the distortion of the infrared images from the optical readout channel. Comparing with the typical optical readout method without any references, the noise equivalent temperature difference (NETD) of such an infrared imaging system with the fiber reference optical readout method can be reduced by about 33% and edges of the IR images become clearer.

Double triangular prism filter based on the optical-readout method in a microelectromechanical infrared imaging system.

This paper presents a novel filtering method with a double triangular prism in an optical-readout thermal imaging system. First, the working principle of this system is described in detail, followed by the analysis of sensitivity. Then, infrared images of hands are obtained. On the basis of the analysis, it is concluded that this filtering method, whose noise equivalent temperature difference (NETD) can reach 145 mK, is effective in obtaining high-quality images. Finally, comparing the filtering method with a knife-edge filter, we can draw the conclusion that the filtering method can effectively improve image quality (the value of NETD is less than that of a knife-edge filter).

[Low-dose Spiral Computed Tomography in Lung Cancer Screening].

Lung cancer is one of the malignant tumors with the highest morbidity and mortality in the world. The low early diagnosis rate and poor prognosis of patients have caused serious social burden. Regular screening of high-risk population by low-dose spiral computed tomography (LDCT) can significantly improve the early diagnosis rate of lung cancer and bring new opportunities for the diagnosis and treatment of lung cancer. In recent years, LDCT lung cancer screening programs have been carried out in many countries around the world and achieved good results, but there are still some controversies in the selection of screening subjects, screening frequency, cost effectiveness and other aspects. In this paper, the key factors of LDCT lung cancer screening, screening effect, pulmonary nodule management and artificial intelligence contribution to the development of LDCT will be reviewed, and the application progress of LDCT in lung cancer screening will be discussed.
.