Blurred Images Research Articles

Dry Eye Subtype Classification Using Videokeratography and Deep Learning.

We previously reported on 'Tear Film Oriented Diagnosis' (TFOD), a method for the dry eye (DE) subtype classification using fluorescein staining and an examination of fluorescein breakup patterns via slit-lamp biomicroscopy. Here, we report 'AI-supported TFOD', a novel non-invasive method for DE subtype classification using videokeratography (VK) and "Blur Value" (BV), a new VK indicator of the extent of blur in Meyer-ring images and deep learning (DL). This study involved 243 eyes of 243 DE cases (23 males and 220 females; mean age: 64.4 ± 13.9 (SD) years)-i.e., 31 severe aqueous-deficient DE (sADDE) cases, 73 mild-to-moderate ADDE (m/mADDE) cases, 84 decreased wettability DE (DWDE) cases, and 55 increased evaporation DE (IEDE) cases diagnosed via the fluorescein-supported TFOD pathway. For DL, a 3D convolutional neural network classification model was used (i.e., the original image and BV data of eyes kept open for 7 s were randomly divided into training data (146 cases) and the test data (97 cases), with the training data increased via data augmentation and corresponding to 2628 cases). Overall, the DE classification accuracy was 78.40%, and the accuracies for the subtypes sADDE, m/mADDE, DWDE, and IEDE were 92.3%, 79.3%, 75.8%, and 72.7%, respectively. 'AI-supported TFOD' may become a useful tool for DE subtype classification.

Three-dimensional thermophotonic super-resolution imaging by spatiotemporal diffusion reversal method.

Spatial super-resolution in thermophotonic imaging was achieved using a combination of spatial second-derivative forming, spatial gradient adaptive filtering, and Richardson-Lucy deconvolution in conjunction with the construction of an experimental point spread function. When implemented through enhanced truncation-correlation photothermal coherence tomography (eTC-PCT), it was possible to restore blurred infrared thermophotonic images to their prediffusion optical resolution state. This modality was tested in various biological applications and proved to be capable of imaging fine axial cracks in human teeth, well-patterned anatomical subsurface structures of a mouse brain, and neovascularization in a mouse thigh due to the rapid proliferation of cancer cells. This modality was found to be immune to optical scattering and could reveal the true spatial extent of biological features at subsurface depths that conventional thermal imaging cannot reach because of limitations imposed by the physics of spreading diffusion.

Single-shot optical imaging with spectrum circuit bridging timescales in high-speed photography.

Single-shot optical imaging based on ultrashort lasers has revealed nonrepetitive processes in subnanosecond timescales beyond the recording range of conventional high-speed cameras. However, nanosecond photography without sacrificing short exposure time and image quality is still missing because of the gap in recordable timescales between ultrafast optical imaging and high-speed electronic cameras. Here, we demonstrate nanosecond photography and ultrawide time-range high-speed photography using a spectrum circuit that produces interval-tunable pulse trains while keeping short pulse durations. We capture a shock wave propagating through a biological cell with a 1.5-ns frame interval and 44-ps exposure time while suppressing image blur. Furthermore, we observe femtosecond laser processing over multiple timescales (25-ps, 2.0-ns, and 1-ms frame intervals), showing that the plasma generated at the picosecond timescale affects subsequent shock wave formation at the nanosecond timescale. Our technique contributes to accumulating data of various fast processes for analysis and to analyzing multi-timescale phenomena as a series of physical processes.

Low-frequency background estimation and noise separation from high-frequency for background and noise subtraction.

In fluorescence microscopy, background blur and noise are two main factors preventing the achievement of high-signal-to-noise ratio (SNR) imaging. Background blur primarily emanates from inherent factors including the spontaneous fluorescence of biological samples and out-of-focus backgrounds, while noise encompasses Gaussian and Poisson noise components. To achieve background blur subtraction and denoising simultaneously, a pioneering algorithm based on low-frequency background estimation and noise separation from high-frequency (LBNH-BNS) is presented, which effectively disentangles noise from the desired signal. Furthermore, it seamlessly integrates low-frequency features derived from background blur estimation, leading to the effective elimination of noise and background blur in wide-field fluorescence images. In comparisons with other state-of-the-art background removal algorithms, LBNH-BNS demonstrates significant advantages in key quantitative metrics such as peak signal-to-noise ratio (PSNR) and manifests substantial visual enhancements. LBNH-BNS holds immense potential for advancing the overall performance and quality of wide-field fluorescence imaging techniques.

<em>Representing the Queen When the Queen Is Missing: The Duchess of Angoulême and Louis XVIII </em> <em>(1799-1824)</em>

This article argues that the Duchess of Angoulême, born Marie-Thérèse Charlotte de France, daughter of King Louis XVI and Queen Marie-Antoinette, took on the role of substitute queen during the reign of her uncle, King Louis XVIII. The King, widower of Marie-Joséphine de Savoie, realised well before his wife’s death the importance of having a female figure by his side. Although the King’s niece was born a Fille de France and did not wear the royal crown, the iconography commissioned by Louis gave her all the prerogatives of queenship, at least in appearance, and so helped shape a blurred and complex image of a royal couple such as had never been seen before.

InertialNet: Inertial Measurement Learning for Simultaneous Localization and Mapping.

SLAM (simultaneous localization and mapping) plays a crucial role in autonomous robot navigation. A challenging aspect of visual SLAM systems is determining the 3D camera orientation of the motion trajectory. In this paper, we introduce an end-to-end network structure, InertialNet, which establishes the correlation between the image sequence and the IMU signals. Our network model is built upon inertial measurement learning and is employed to predict the camera's general motion pose. By incorporating an optical flow substructure, InertialNet is independent of the appearance of training sets and can be adapted to new environments. It maintains stable predictions even in the presence of image blur, changes in illumination, and low-texture scenes. In our experiments, we evaluated InertialNet on the public EuRoC dataset and our dataset, demonstrating its feasibility with faster training convergence and fewer model parameters for inertial measurement prediction.

Photoacoustic image reconstruction with anobjective function using TGV and ESTGV asaregularization functional.

Photoacoustic tomographic imaging is a non-invasive medical diagnostic technology for visualizing biological tissue. However, the inverse problem and noise in photoacoustic signals often cause blurred images. Existing regularization methods struggle with staircasing artifacts and edge preservation. To overcome this, an objective function incorporating total generalized variation (TGV) is proposed. However, it failed with high-density Gaussian noise. To address this, an extended version called edge-guided second-order TGV (ESTGV) is introduced. For sparsification, wavelet transform and discrete cosine transform are introduced, while the fast-composite-splitting algorithm is employed for the inverse problem solution. Experimental validation demonstrates the potential of these approaches.

Hierarchical deconvolution dehazing method based on transmission map segmentation.

Images captured in fog are often affected by scattering. Due to the absorption and scattering of light by aerosols and water droplets, the image quality will be seriously degraded. The specific manifests are brightness decrease, contrast decrease, image blur, and noise increase. In the single-image dehazing method, the image degradation model is essential. In this paper, an effective image degradation model is proposed, in which the hierarchical deconvolution strategy based on transmission map segmentation can effectively improve the accuracy of image restoration. Specifically, the transmission map is obtained by using the dark channel prior (DCP) method, then the transmission histogram is fitted. The next step is to divide the image region according to the fitting results. Furthermore, to more accurately recover images of complex objects with a large depth of field, different levels of inverse convolution are adopted for different regions. Finally, the sub-images of different regions are fused to get the dehazing image. We tested the proposed method using synthetic fog images and natural fog images respectively. The proposed method is compared with eight advanced image dehazing methods on quantitative rating indexes such as peak signal-to-noise ratio (PSNR), structural similarity (SSIM), image entropy, natural image quality evaluator (NIQE), and blind/referenceless image spatial quality evaluator (BRISQUE). Both subjective and objective evaluations show that the proposed method achieves competitive results.

Restoration and enhancement optimization of blurred images based on SRGAN

Blurred images pose a significant challenge in many applications, including medical imaging, remote sensing, and surveillance systems. These images suffer from low resolution, noise, and missing data, which can hinder their interpretation and analysis. Traditional methods for image restoration and enhancement have their limitations, such as low quality and slow processing times. To overcome these challenges, this paper proposes an innovative method using Super-resolution Generative Adversarial Networks (SRGANs) to enhance image quality and fidelity. The proposed method employs adversarial training, perceptual loss, residual learning, and feature reconstruction to generate visually realistic and high-quality super-resolution (SR) images from low-resolution (LR) inputs. The SRGANs approach outperforms traditional methods, demonstrating its potential to advance image restoration and enhancement techniques. The paper also discusses possible improvements and future directions for this technique.

Right middle pulmonary vein classification by 3D-CT reconstruction and its clinical significance.

A profound understanding of the type of right middle lobe (RML) vein return is crucial for thoracic surgeons to ensure safe anatomic lung resection. In this study, the type of venous return in the RML and its clinical significance were analyzed using the 3D computed tomography (3D-CT) reconstruction technique. This retrospective study analyzed the patients who underwent anatomical lobectomy or anatomical lung segment resection with preoperative 3D-CT reconstruction at the Department of Thoracic Oncology Surgery, Fujian Cancer Hospital from January 2020 to October 2022. The exclusion criteria were as follows: (I) blurred CT images and inadequate IQQA 3D reconstruction for clear visualization of bronchial and vascular subsegmental branches; (II) patients with a previous history of right middle lobectomy; (III) incomplete medical records. This study included 608 eligible cases, 245 males (40.30%) and 363 females (59.70%), with an average age of 58.61 years. We identified four major types of RML venous return, including 407 cases (66.94%) of central vein type, 123 cases (20.23%) of isolated vein type, 11 cases (1.81%) of basal vein type, and 67 cases (11.02%) of combined vein type. Furthermore, the central venous type comprised four subtypes: 191 cases (27.3%) had one branch, 165 cases (27.14%) had two branches, 43 cases (7.07%) had three branches, and 8 cases (1.32%) had more than three branches. The combined venous type includes four subtypes: combination type A (n=37 cases, 6.09%), combination type B (n=7 cases, 1.15%), combination type C (n=21 cases, 3.45%), and combination type D (n=2 cases, 0.33%). Based on 3D-CT reconstruction, numerous types of venous return in the RML were found, highlighting the significance of 3D-CT reconstruction for thoracic surgeons. It plays a crucial role in understanding the type of RML venous return prior to surgery, providing invaluable information to ensure surgical safety and success.

AFA-Net: Adaptive Feature Attention Network in image deblurring and super-resolution for improving license plate recognition

Although a number of license plate recognition (LPR) systems have become significantly advanced, they are still far from producing ideal images. They are still fiddled with low-resolution (LR) and motion blur, which are two of the most common problems in images extracted from automobile driving environment. In order to address this issue, we present a novel LPR method that processes LR and motion blurred images from dash cams. We propose a unique framework, AFA-Net (Adaptive Feature Attention Network) organized by image pre-restoration, feature composition and image reconstruction modules. The proposed AFA-Net can generate a clear restoration image for robust LPR performance with the images obtained from dash cams in an unconstrained environment. Furthermore, we explore the novel problem, Joint-IRLPRNet (Joint-Image Restoration and License Plate Recognition Network), that simultaneously address image restoration (i.e. SR and deblurring) and LPR in an end-to end trainable manner. Moreover, we introduce a dataset called LBLP (LR and blurred license plate (LP)). The dataset is composed of 2,779 LR and motion blurred cropped LP images, extracted from unconstrained dash cams. The experimental results on LBLP dataset indicate that AFA-Net achieves 15.28% improvement in recognition accuracy, 6.47% in sequence similarity, and 3.89% in character similarity, compared to the traditional LPR model with image restoration model. Moreover, Joint-IRLPRNet can be more effective results than AFA-Net.

The Elusive Nature of “Seeing”

Atmospheric image blur, “seeing”, is one of the key parameters that influences the selection of observatory sites and the performance of ground-based telescopes. In this review, the common definition of seeing based on the Kolmogorov turbulence model is recalled. The ability of this model to represent real, non-stationary fluctuations of the air refractive index is discussed. Even in principle, seeing (a model parameter) cannot be measured with arbitrary accuracy; consequently, describing atmospheric blur by a single number, seeing, is a crude approximation. The operating principles of current seeing monitors are outlined. They measure optical effects caused by turbulence, sampling certain regions of spatial and temporal spectrum of atmosphreic optical disturbances, and interpret their statistics in the framework of the standard model. Biases of seeing monitors (measurement noise, propagation, finite exposure time, optical defects, wind shake, etc.) should be quantified and corrected using simulations, while instrument comparison campaigns serve as a check. The elusive nature of seeing follows from its uniqueness (a given measurement cannot be repeated or checked later), its non-stationarity (dependence on time, location, and viewing direction), a substantial role of the highly variable surface layer, and a potential bias caused by the air flow in the immediate vicinity of the seeing monitors. The results of seeing measurements are outside the scope of this review.

Ex uno plures: how to construct high-speed movies of collapsing cavitation bubbles from a single image

The time-resolved visualization of the dynamics of a cavitation bubble usually requires the use of expensive high-speed cameras, which often provide a limited spatial resolution. In the present study, we propose an alternative to these high-speed imaging techniques. The method is based on the recently introduced virtual frame technique, which relates the motion of a monotonic propagating front to the resulting image blur captured on a long-exposure shadowgraph. We use a consumer-level camera to photograph the entire collapse phase of cavitation bubbles. We then demonstrate that both the dynamics of a spherically collapsing bubble and those of a bubble collapsing near a rigid boundary can be accurately reconstructed from this single photograph at a virtual frame rate of up to 2 Mfps on a 24.2 Mpx sensor.

MFENet: Multi-scale feature extraction network for images deblurring and segmentation of swinging wolfberry branch

Wolfberry harvesting faces challenges related to factors such as wind-induced branch movement and mechanical arm collisions, which lead to undesirable swinging of wolfberry branches. This motion results in blurred images captured by imaging devices, which hampers the effectiveness of the harvesting robot arm in identifying and grasping the wolfberry branches. Clear, unblurred images are imperative for precise branch recognition and successful harvesting. Unfortunately, prior research on wolfberry branch identification has largely overlooked the issue of motion blur. In this study, we propose a novel solution by introducing the Multi-Scale Feature Extraction Network (MFENet) specifically tailored for image deblurring tasks. Our approach aims to enhance the accuracy of identifying and tracking dynamically moving wolfberry branches. MFENet achieves this by harnessing a combination of multi-scale image content features, incorporating modules for multi-scale depth-separable convolution, multi-scale receptive field structure, and attention refinement residual processing. In this study, we propose a novel solution by introducing the Multi-Scale Feature Extraction Network (MFENet) specifically tailored for image deblurring tasks. Our approach aims to enhance the accuracy of identifying and tracking dynamically moving wolfberry branches. MFENet achieves this by harnessing a combination of multi-scale image content features, incorporating modules for multi-scale depth-separable convolution, multi-scale receptive field structure, and attention refinement residual processing. Our findings underscore the efficacy of the proposed MFENet in addressing the challenge of blurred images resulting from the swinging motion of wolfberry branches. Furthermore, through experimental validation, we demonstrate that the improved SparseInst segmentation method achieves exceptional precision and faster segmentation speeds when applied to the branch dataset. This research not only provides a practical solution to mitigate branch swinging during wolfberry harvesting but also offers valuable theoretical and technical support to advance this field.

Underwater visual feature matching based on attenuation invariance

Underwater visual feature matching plays an important role in the localization and autonomous motion of underwater robots. However, feature descriptors are prone to change in degraded underwater images due to light attenuation, which leads to image color distortion, blur, and low illumination. Although the recent developments in deep learning have improved the performance of feature descriptors, the degradation of underwater images and the lack of underwater datasets have hindered the development of underwater visual feature matching and localization techniques. Herein, we propose an underwater visual feature matching method based on attenuation invariance. Our approach involves two main components. First, we study the attenuation differences between images to generate feature descriptors that are invariant to attenuation, thereby improving the accuracy of underwater visual feature matching. Second, we construct two datasets: the multiple water type (MWT) dataset, which contains over 30000 underwater images, and the underwater image feature descriptor (UIFD) evaluation dataset. These datasets are developed to compensate for the lack of evaluation datasets for underwater visual feature matching. The code and datasets are available at https://github.com/Sun-Jinghao/UIFD.

Contrast Adaptation in Pseudophakic Patients with Macular Disorders

Purpose To learn whether contrast adaptation, induced by positive spherical defocus, is compromised by macular disorders such as age-related macular degeneration (AMD) or epiretinal membranes (ERM) and to gain further insight in the functionality of the pathological macula and the level of “functional reserve” often postulated for the indication of presbyopia correcting IOLs. Methods In a pilot study, patients with macular disorders, AMD and ERM, (n = 10) and healthy volunteers (n = 10) were tested to quantify contrast adaption after +4 D defocus for 10 min, by performing an interocular contrast matching task. Subjects manually adjusted the perceived contrast of the test patch as seen by the test eye to match to the contrast of a target patch with a fixed Michelson contrast of 0.2 as seen by the contralateral untreated eye. Results Patients with macular disorders subjectively matched the 0.2 target contrast with a contrast of 0.24 ± 0.06 (mean ± SD) before adaptation and with a contrast of 0.19 ± 0.04 after adaptation (p < 0.05). Accordingly, patients with macular disorders showed an induced contrast gain by adaptation of 0.05 (27%), which was not different from the control group, which showed an induced contrast gain by adaptation of 0.06 (35%). Conclusion Patients with mild and moderate macular disorders, such as AMD and ERM, show an induced contrast adaptation, i.e. a gain in contrast sensitivity, at 3.2 cpd, which is not different in level from the induced contrast adaptation in healthy subjects. Macular disorders do not prevent adaptation of the patient’s visual system to low contrast or blurred retinal images. Therefore, the implantation of presbyopia correcting IOLs is not a strict exclusion criterion for these patients, but the progressive nature of the macular disorder must be taken into account.

A Motion Deblurring Network for Enhancing UAV Image Quality in Bridge Inspection

Unmanned aerial vehicles (UAVs) have been increasingly utilized for facility safety inspections due to their superior safety, cost effectiveness, and inspection accuracy compared to traditional manpower-based methods. High-resolution images captured by UAVs directly contribute to identifying and quantifying structural defects on facility exteriors, making image quality a critical factor in achieving accurate results. However, motion blur induced by external factors such as vibration, low light conditions, and wind during UAV operation significantly degrades image quality, leading to inaccurate defect detection and quantification. To address this issue, this research proposes a deblurring network using a Generative Adversarial Network (GAN) to eliminate the motion blur effect in UAV images. The GAN-based motion deblur network represents an image inpainting method that leverages generative models to correct blurry artifacts, thereby generating clear images. Unlike previous studies, this proposed approach incorporates deblur and blur learning modules to realistically generate blur images required for training the generative models. The UAV images processed using the motion deblur network are evaluated using a quality assessment method based on local blur map and other well-known image quality assessment (IQA) metrics. Moreover, in the experiment of crack detection utilizing the object detection system, improved detection results are observed when using enhanced images. Overall, this research contributes to improving the quality and accuracy of facility safety inspections conducted with UAV-based inspections by effectively addressing the challenges associated with motion blur effects in UAV-captured images.

Research on environmental adaptability of AI-based visual perception system under the perspective of vibration

Artificial intelligence (AI) models are widely used in intelligent devices. However, challenging vibration environments can seriously affect the environmental adaptability of AI-based devices. In this paper, an environmental adaptability assessment method of AI-based devices is proposed. The formation mechanism of optical blurred image is elaborated, and measure function is used to transform the stress magnitudes of environment into distances of distribution discrepancy. To ensure that the obtained optical blurred images are fully controllable, we generate vibration-blurred images using an explicitly quantifiable extended dataset. Additionally, the generated vibration-blurred image is evaluated for consistency with the actual image. Finally, the environmental adaptability of multiple state-of-the-art models is evaluated based on extended datasets. The assessment results show that the environmental adaptability of the visual perception system is mainly limited by the vibration amplitude under high frequency. Under low frequency vibration, the environmental adaptability of AI models is also frequency-dependent.

Comparison of improved variational autoencoder models for human face generation

The variational autoencoder (VAE) model has evolved a number of VAE improved models in the past 10 years, including CVAE, WAE, NVAE, etc. These models have greatly improved the calculation speed of VAE and the resolution of generated images. The main goal of this paper is to compare the principles of these different models and the effect of generating images. The principle analysis method is mainly used to study the improvement direction of the VAE model in different papers. The main ideas for improving the VAE model include optimizing the loss function, optimizing the objective function, introducing other parameters, or improving code efficiency. Some models even add many algorithms in computer vision in for improving image effect. In the experiments in this paper, results show the image results processed by different models. In the current optimal model NVAE, this algorithm has solved most of the VAE image blur problems, and has achieved perfection in image details. The improvement of the VAE model in the future may require innovative ideas. Under the current principle and model structure, the room for improvement of the VAE model is relatively limited. The final experiment shows that to improve the quality of VAE generated images, it is required to optimize the objective function, optimize the algorithm and add image visual processing, and then the generated images will be significantly improved.

Multi-polarization selective absorber based on terahertz metasurface

Tunable absorbers attract much attention due to their wide applications, such as integrated optical sensors, multiplexing far-field imaging and so on. In this paper, we design a multi-polarization selective absorber based on terahertz metasurface, whose geometric symmetry is tunable to satisfy the requirements of linear and circular polarization selective absorbers. At the ON state, the metasurface selectively absorbs y-polarized wave at 1.0 THz, with the linear dichroism of 0.84. At the OFF state, the metasurface selectively absorbs left circularly polarized wave at 1.16 THz, with the circular dichroism of 0.81. The physical mechanism of the metasurface is explained by the Jones calculus and the electromagnetic field distributions. In addition, the multi-polarization selective behavior is insensitive to the incident angle. Finally, we design a sample to verify the potential application of the metasurface absorber, which displays multiplexing imaging of ‘C’, ‘mirrored C’, ‘O’ and blurred images under different polarized incidences at different states. Such a metasurface, switching between linear polarization selective absorber and circular polarization selective absorber, may achieve stable applications in multifunction integrated absorbers, detectors, and image encryptions.