Search

Existing deep unfolding methods unroll an optimization algorithm with a fixed number of steps, and utilize convolutional neural networks (CNNs) to learn data-driven priors. However, their performance is limited for two main reasons. Firstly, priors learned in deep feature space need to be converted to the image space at each iteration step, which limits the depth of CNNs and prevents CNNs from exploiting contextual information. Secondly, existing methods only learn deep priors at the single full-resolution scale, so ignore the benefits of multi-scale context in dealing with high level noise. To address these issues, we explicitly consider the image denoising process in the deep feature space and propose the deep unfolding multi-scale regularizer network (DUMRN) for image denoising. The core of DUMRN is the feature-based denoising module (FDM) that directly removes noise in the deep feature space. In each FDM, we construct a multi-scale regularizer block to learn deep prior information from multi-resolution features. We build the DUMRN by stacking a sequence of FDMs and train it in an end-to-end manner. Experimental results on synthetic and real-world benchmarks demonstrate that DUMRN performs favorably compared to state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Low illumination, under which discriminative clues are buried in the captured images, is an under-investigated but noteworthy issue in wild scene text detection. Existing deep learning approaches suffer from the scarcity of training data and illumination-sensitive feature representation. To address these issues, we propose a Low Illumination Scene Text (LIST) Detector training with authentic synthetic data and integrating dedicated feature enhancement modules. Specifically, we adopt a lightweight and non-reference low-light scene text image synthesis network to acquire adequate training data through pixel-wisely adjusting the dynamic range curve of normal-light images. Moreover, illumination invariant feature representation is learned through dual path feature extraction stem with intensity adjusted inputs, and feature fusion branch with automatically designed fusion cell. In the end, the enhanced feature is fed into a segmentation-based layer to localize arbitrary shape text instances. We construct a labeled real-world scene text image dataset called "DarkText" and conduct extensive experiments to validate the advantages of our proposed framework over state-of-the-art competitors. [ABSTRACT FROM AUTHOR]

Fast and accurate semantic analysis of natural disaster images is crucial for rational rescue plans and resource allocation. However, the scarcity of meticulously labelled datasets and the ignorance of region-of-interest scale variations of popular general-purpose methods lead to undesirable performance. In this paper, we propose a novel triple-strip attention mechanism (TSAM) to solve the generalization problem of disaster images that can be combined into general networks. Our TSAM accumulates features of three parallel-strip attentions (row strip attention, column strip attention, and channel strip attention), and the output is multiplied with original input features for further processing. Our attention mechanism can effectively overcome the defect of ignoring global features caused by the convolution and enhance the performance of the network by weighting the features from both spatial and channel aspects more comprehensively. Besides, we employ both the compression and expansion operations in the weighting operation to reduce the amount of parameters, leading to negligible computational overhead. Experiments validate that our TSAM outperforms other state-of-the-art methods on natural disaster segmentation. Due to its concise form, plug-and-play pattern, and high promotion rate, our TSAM can be combined with many existing neural networks for better performance improvement. [ABSTRACT FROM AUTHOR]

Similarity graph learning is the most key technique for multiview spectral clustering. However, existing methods fail when applied to uncertain data contaminated with various types of noise in an open environment. Due to the damaged structure by noise, unreliable similar relationships are learned, which extends similarity inconsistency among views. Moreover, the high-order correlation hidden in graphs are ignored generally. To address these problems, we propose a reliable similarity learning scheme for multiview clustering on uncertain data. This method can significantly improve spectral clustering performance in a noisy environment, and the contributions of our scheme include the following three aspects: 1) Uncertain data subspace reconstruction and adaptive graph learning are combined to construct a view-specific graph from high-quality recovered data, thus improving robustness. 2) A low-rank tensor constraint is utilized to facilitate multiview fusion, where the latent high-order correlation among view graphs will be fully explored when learning the consensus graph structure. 3) Data recovery, view-specific graphs, and latent consensus tensor structure are assembled into a unified framework, to be optimized jointly for mutual benefit. Our study also develops an efficient algorithm for obtaining overall solutions. The experimental results on several datasets demonstrate that our proposed approach shows significant improvements in robustness and evaluation metrics over the comparison methods. [ABSTRACT FROM AUTHOR]

We propose a framework to generate customized summarizations of visual data collections, such as collections of images, materials, 3D shapes, and 3D scenes. We assume that the elements in the visual data collections can be mapped to a set of vectors in a feature space, in which a fitness score for each element can be defined, and we pose the problem of customized summarizations as selecting a subset of these elements. We first describe the design choices a user should be able to specify for modeling customized summarizations and propose a corresponding user interface. We then formulate the problem as a constrained optimization problem with binary variables and propose a practical and fast algorithm based on the alternating direction method of multipliers (ADMM). Our results show that our problem formulation enables a wide variety of customized summarizations, and that our solver is both significantly faster than state‐of‐the‐art commercial integer programming solvers and produces better solutions than fast relaxation‐based solvers. [ABSTRACT FROM AUTHOR]

Herein, a series of phosphorus-doped CoMoO4 rod bundles electrocatalysts on 3D nickel foams have been successfully synthesized and exploited as efficient hydrogen evolution reaction (HER) electrocatalyst in 1 M KOH. The incorporation of phosphorus into CoMoO4 can significantly promote the HER activity, and the CoMoO4 bundles phosphorized at 350 °C exhibited the optimum activity, which only required low overpotentials of 56 and 148 mV to deliver cathodic current densities of 10 mA cm−2 and 100 mA cm−2, respectively. This electrode also presented considerable long-term electrochemical stability with negligible delay after 30 h operation and 3000 accelerated cyclic voltammetry cycles. Our study provides a superior earth-abundant catalyst for efficient HER in alkaline media, which also suggests that P doping engineering is an effective way to boost the HER activity of transition metal-based oxides. [ABSTRACT FROM AUTHOR]

Rational design of high efficient and low cost oxygen evolution reaction (OER) electrodes is critical for electrochemical water splitting. Herein, we report Reevesite Ni−Fe layered double hydroxide (NiFe‐(CO32−)‐LDH) nanosheet arrays as novel OER catalysts in alkaline electrolyte. These self‐supported nanosheet arrays with a composition of Ni6Fe2(CO3)(OH)16 ⋅ 4H2O on Ni foam (NF) were synthesized by a one‐pot hydrothermal method, in which NF served as both the conductive substrate and Ni source. The morphology and OER activity can be tuned by iron concentration. In 1 M KOH, the optimized NiFe‐(CO32−)‐LDH catalyst only required overpotentials of 228 and 264 mV to deliver current densities of 10 and 100 mA cm−2, respectively, with a small Tafel slope of 37 mV dec−1. Remarkably, this catalyst also showed outstanding durability by long‐term operation for 20 h with 96 % retention. Our work provides attractive cost‐effective 3D catalysts for efficient and robust water oxidation at alkaline media. [ABSTRACT FROM AUTHOR]

Inspired by the knowledge that a nonchiral metal nanocrystal could be bestowed with chirality when the metal nanocrystal and the chiral molecules form complexes, in this contribution, a two-step pyrolytic route to obtain chiroptical carbon quantum dots (CQDs) was attempted through the optically active surface passivation procedure with enantiomers. Here, tartaric acid (TA) and citric acid (CA), one being a chiral molecule and the other being an achiral molecule, were used as carbon sources to produce a carbon core, respectively, and D / L -penicillamine (D / L -Pen) molecules were utilized as chiral ligands to further passivate it in the two-step pyrolytic route. The strategy is applicative no matter whether the carbon source is from chiral or achiral molecules during the first carbonization step. Also, the second pyrolytic reaction temperature, lower than the melting point of the surface passivation enantiomeric ligand, is critical to preserve the chiral information and to generate the chiral centers of the CQDs. Based on this simple approach, it is possible to envision the design and fabrication of new types of more sophisticated optically active photoluminescence carbon nanodots, which may be applied in chiral catalysis, chiral recognition and sensing. [ABSTRACT FROM AUTHOR]

Biological thiols play a key role in biological processes and are involved in a variety of diseases. The discriminative detection of biological thiols is still challenging. In this contribution, a platform, an energy transfer-based quenching system composed of nitrogen and sulphur co-doped carbon dots (N, S-CDs) and gold nanoparticles (AuNPs), was established to discriminate glutathione (GSH) from other competitive biothiols including cysteine (Cys) and homocysteine (Hcy) based on a photoluminescence (PL) “switch-on” signal readout. The presence of GSH can encapsulate AuNPs in priority because of the strong affinity towards AuNPs and the steric hindrance effect of GSH, leaving little chance for the N, S-CDs binding on the surface of AuNPs and thus resulting in the PL recovery of N, S-CDs. Compared with the nitrogen-doped carbon dots (N-CDs), the N, S-CDs can enhance 10 times sensitivity for the designed PL “switch-on” sensing strategy. The proposed method has a detection limit of 3.6 nM and can be successfully applied for the detection of GSH in human serum. [ABSTRACT FROM AUTHOR]

To investigate in patients with retinoblastoma (RB). We collected 0.1 ml AH of 35 children with RB and 20 patients with congenital cataracts as controls. Multiplex enzyme-linked immunosorbent assays (ELISAs) and Luminex xMAP technology were used to assess 45 cytokines/chemokines, matrix metalloproteinases (MMPs), and acute-phase proteins in the identification cohort. The concentrations of IL-6, IL-7, IL-8, IFN-γ, PIGF-1, VEGF-A, β-NGF, HGF, EGF and FGF-2 were significantly higher in the AH of patients with RB than those in the control group (P<0.05). The study showed that the higher levels of IP-10, IL-6, IL-7, IL-8, IFN-γ, PIGF-1, VEGF-A, β-NGF, HGF, EGF and FGF-2 in AH may be associated with RB. Our findings may facilitate a better understanding of the molecular pathways of tumors and solid molecular targets for new strategies for therapy and the earlier diagnosis of RB. [ABSTRACT FROM AUTHOR]

The bilateral filter (BF) has showed great effectiveness for a variety of problems. However, its brute‐force implementation is time consuming. One way of accelerating a BF is to approximate the nonlinear range kernel of the BF by a set of linear time shiftable kernels. To achieve this goal, only finite values of the kernel of the BF have been used to perform smoothing due to the quantisation of digital images. Thus, the filtering results are not changed by substituting the range kernel with the function having the same values at finite discrete points. The Lagrange interpolation polynomial can exactly pass through predefined points and therefore can be employed to replace original kernels for accurate by accelerating the BF. To speed up the BF at the cost of small approximation error, two approximation methods are proposed to obtain the optimal fitting polynomial. The performance of the proposed method is validated by extensive experiments. [ABSTRACT FROM AUTHOR]

The bilateral filter (BF) as an edge‐preserving lowpass filter is a valuable tool in various image processing tasks, including noise reduction and dynamic range compression. However, its computational cost is too high to apply in the real‐time processing tasks as the range kernel, which acts on the pixel intensities, making the averaging process nonlinear and computationally intensive, particularly when the spatial filter is large. Using the well‐known Hermite polynomials, a BF accelerating method is proposed, which reduces the computational complexity from O(r2n) to O(n), where r denotes the filter size of a BF and n is the total number of pixels in an image. [ABSTRACT FROM AUTHOR]

Due to the heavy computation cost, the bilateral filter cannot be applied to real‐time applications. To deal with this problem, many range kernel approximation based acceleration schemes have been proposed. However, all these methods use predefined basis function and assign the same penalisation for all range values to fulfil range kernel approximation. These constraints block from further improving the filtering accuracy and speed. Employing weighted variable projection, a new acceleration method which achieves state‐of‐the‐art performance is proposed. This is done by: (i) unleashing the constraint of using fixed basis function; (ii) exploiting the colour distribution information of the input image to perform a weighted approximation of the range kernel. Experiments demonstrate the superiority of the proposed method in gaining more accurate filtering results efficiently. [ABSTRACT FROM AUTHOR]