Search

Single image super-resolution (SISR) using deep convolutional neural networks (CNNs) achieves the state-of-the-art performance. Most existing SISR models mainly focus on pursuing high peak signal-to-noise ratio (PSNR) and neglect textures and details. As a result, the recovered images are often perceptually unpleasant. To address this issue, in this paper, we propose a texture and detail-preserving network (TDPN), which focuses not only on local region feature recovery but also on preserving textures and details. Specifically, the high-resolution image is recovered from its corresponding low-resolution input in two branches. First, a multi-reception field based branch is designed to let the network fully learn local region features by adaptively selecting local region features in different reception fields. Then, a texture and detail-learning branch supervised by the textures and details decomposed from the ground-truth high resolution image is proposed to provide additional textures and details for the super-resolution process to improve the perceptual quality. Finally, we introduce a gradient loss into the SISR field and define a novel hybrid loss to strengthen boundary information recovery and to avoid overly smooth boundary in the final recovered high-resolution image caused by using only the MAE loss. More importantly, the proposed method is model-agnostic, which can be applied to most off-the-shelf SISR networks. The experimental results on public datasets demonstrate the superiority of our TDPN on most state-of-the-art SISR methods in PSNR, SSIM and perceptual quality. We will share our code on https://github.com/tocaiqing/TDPN. [ABSTRACT FROM AUTHOR]

Convolutional Neural Networks (CNN) are widely used in image dehazing. However, existing network frameworks are built based on manual design from practical experience, lacking interpretable result or theoretical guidelines. Recently, residual networks are regarded as the explicit Euler forward approximation of the ODE (Ordinary Differential Equation), and several ODE-inspired networks are proposed based on the low-order explicit Euler schemes. However, on the issues of system stability and training convergence, high-order Implicit Adams Predictor–Corrector (IAPC) methods have proven to be better than low-order explicit Euler methods. Hence, we extend the IAPC method to the High-order Implicit Adams Network (HIAN). To do so, we design a series of Implicit Adams Predictor–Corrector Blocks (IABs) based on the high-order IAPC methods, all of which give better stability and accuracy than the ones designed using the low-order Euler methods. Given that, we further propose the Implicit Adams Predictor–Corrector Module (IAM) by combining the Non-local Sparse Attention (NSA) and Attention Feature Fusion (AFF) with stacked IABs where the NSA explores the mutual-correlation among intermediate features with low computation cost via a sparse constraint, while the AFF fuses intermediate features by reweighting the features from stacked IABs adaptively. Moreover, because manual network design with IABs limits dehazing performance, the Neural Architecture Search (NAS) is used to find an optimal architecture automatically. This resulting design not only is interpretable for image dehazing but also provides a reliable guideline on future network designs. The experiments demonstrate that the proposed method outperforms most existing methods on both synthetic and real images. • Present a high-order Implicit Adams Network using Neural Architecture Search for image dehazing. • Design high-order Implicit Adams Predictor–Corrector (IAPC) blocks inspired by the corresponding IAPC methods. • Develop IAPC modules by taking advantage of Non-local Sparse Attention, the Attention Feature Fusion and IAPC blocks. [ABSTRACT FROM AUTHOR]

In recent years, disparity estimation of a scene based on deep learning methods has been extensively studied and significant progress has been made. In contrast, a traditional image disparity estimation method requires considerable resources and consumes much time in processes such as stereo matching and 3D reconstruction. At present, most deep learning based disparity estimation methods focus on estimating disparity based on monocular images. Motivated by the results of traditional methods that multi-view methods are more accurate than monocular methods, especially for scenes that are textureless and have thin structures, in this paper, we present MDEAN, a new deep convolutional neural network to estimate disparity using multi-view images with an asymmetric encoder–decoder network structure. First, our method takes an arbitrary number of multi-view images as input. Next, we use these images to produce a set of plane-sweep cost volumes, which are combined to compute a high quality disparity map using an end-to-end asymmetric network. The results show that our method performs better than state-of-the-art methods, in particular, for outdoor scenes with the sky, flat surfaces and buildings. [ABSTRACT FROM AUTHOR]

• The proposed network contains an attentive recurrent network and a U-recurrent encoder-decoder network. • The attentive recurrent network generates a haze attention map for highlighting the haze regions. • The U-recurrent encoder-decoder network predicts the clear image with the guidance of the haze attention map. • A new residual pyramid pooling module enlarges the receptive field of the whole network. Haze removal is an important pre-processing step in many computer vision tasks. Convolutional neural networks, especially the U-shaped networks, have shown to be effective in image dehazing. Nevertheless, these networks have three main limitations. First, the relevant haze information, e.g. concentration of haze, is totally ignored. Second, spatial inconsistency and information dilution usually occur when the networks refine the dehazed results with a coarse-to-fine strategy. Third, the receptive field of the network is not large enough to capture structural information. Motivated by these problems, a new attentive U-recurrent encoder-decoder dehazing network is presented, which consists of an attentive recurrent network and a U-recurrent encoder-decoder network. By assuming that haze layers with different depths can be detected by multiple stages, we use an attentive recurrent network to generate the haze attention map for guiding the U-recurrent encoder-decoder network with the concentration of haze to better estimate the clear image. Meanwhile, the features for dehazing are further enhanced and the dehazing results are refined in the U-recurrent encoder-decoder network. This design not only enables spatial consistency but also reduces information dilution with short recurrent pathways. Furthermore, a novel residual pyramid pooling module is also proposed and used in the U-recurrent encoder-decoder network, which provides the network with structural information and with an enlarged receptive field. The experimental results demonstrate that our method outperforms state-of-the-art dehazing algorithms on both synthetic and real hazy images. [ABSTRACT FROM AUTHOR]

Recent developments in Convolutional Neural Networks (CNNs) have made them one of the most powerful image dehazing methods. In particular, the Residual Networks (ResNets), which can avoid the vanishing gradient problem effectively, are widely deployed. To understand the success of ResNets, recent mathematical analysis of ResNets reveals that a ResNet has a similar formulation as the Euler method in solving the Ordinary Differential Equations (ODE's). Hence, image dehazing which can be formulated as an optimal control problem in dynamical systems can be solved by a single-step optimal control method, such as the Euler method. This optimal control viewpoint provides a new perspective to address the problem of image restoration. Motivated by the advantages of multi-step optimal control solvers in ODE's, which include better stability and efficiency than single-step solvers, e.g. Euler, we propose the Adams-based Hierarchical Feature Fusion Network (AHFFN) for image dehazing with modules inspired by a multi-step optimal control method named the Adams–Bashforth method. Firstly, we extend a multi-step Adams–Bashforth method to the corresponding Adams block, which achieves a higher accuracy than that of single-step solvers because of its more effective use of intermediate results. Then, we stack multiple Adams blocks to mimic the discrete approximation process of an optimal control in a dynamical system. To improve the results, the hierarchical features from stacked Adams blocks are fully used by combining Hierarchical Feature Fusion (HFF) and Lightweight Spatial Attention (LSA) with Adams blocks to form a new Adams module. Finally, we not only use HFF and LSA to fuse features, but also highlight important spatial information in each Adams module for estimating the clear image. The experimental results using synthetic and real images demonstrate that the proposed AHFFN obtains better accuracy and visual results than that of state-of-the-art methods. • We propose the Adams-based Hierarchical Feature Fusion Network (AHFFN) for image dehazing. • The Adams Module is designed based on Adams Blocks, inspired by multi-step high-order Adams–Bashforth methods. • HFF and LSA are proposed to fuse features and pay attention to heavy haze pixels within and among the Adams Module. [ABSTRACT FROM AUTHOR]

Image dehazing is an improtant image pre-processing step for many computer vision tasks with many proposed methods using convolutional neural networks. Ordinary Differential Equations (ODE) are a powerful mathematical tool in social and natural science, leading to better understanding, prediction, and use of information systems. Because prior knowledge has demonstrated its effectiveness in many practical domains, we design a Visual Attention Network (VAN) and an ODE-inspired Network (ODEN) based on prior knowledge. Then, we develop a multi-model fuzzy fusion strategy, which integrates results predicted by the prior-based Visual Attention Network (VAN) and the ODE-inspired Network (ODEN) into a single network to leverage their respective strengths in improving the dehazing performance. First, to utilize the haze-related prior for dehazing, the VAN removes haze with the help of the haze attention map. Then, to improve performance of stacked residual blocks inspired by the first-order Euler method in ODE-based methods, the ODEN is built by only 3 Runge–Kutta Modules (RKM), each of which not only is related to the stable fourth-order Runge–Kutta method implemented in the Runge–Kutta Block (RKB), but also combines the RKB with an attention mechanism. Finally, an attention based fusion mechanism is used to fuse results estimated by the ODEN and the VAN based on the multi-level features extracted by a pretrained ResNeXt. The experimental results demonstrate that the proposed method outperforms existing state-of-the-art methods in terms of visual effects and accuracy. The average PSNR and SSIM on three public datasets are 31.36 and 0.9766, respectively, which are better than the compared state-of-the-art methods. [ABSTRACT FROM AUTHOR]

• We proposed a novel method termed the "PKG-Net" to accurately predict Alzheimer's disease. • The input is designed to be multi-dimensional and collaboratively represents lesion area from multiple scales via the pyramid representation. • The joint loss function is utilized to improves the practicability of the proposed network and its stability in training. • Our method demonstrates good generalization ability, and has achieved excellent results of 97.28% in accuracy on the ADNI dataset. Recently, deep learning based Computer-Aided Diagnosis methods have been widely utilized due to their highly effective diagnosis of patients. Although Convolutional Neural Networks (CNNs) are capable of extracting the latent structural characteristics of dementia and of capturing the changes of brain anatomy in Magnetic Resonance Imaging (MRI) scans, the high-dimensional input to a deep CNN usually makes the network difficult to train, and affects its diagnostic accuracy. In this paper, a novel method called the hierarchical pseudo-3D convolution neural network based on a kernel attention mechanism with a new global context block, which is abbreviated as "PKG-Net", is proposed to accurately predict Alzheimer's disease even when the input features are complex. Specifically, the proposed network first extracts multi-scale features from pre-processed images. Second, the attention mechanism and global context blocks are applied to combine features from different layers to hierarchically transform the MRI into more compact high-level features. Then, a joint loss function is used to train the proposed network to generate more distinguishing features, which improve the generalization performance of the network. In addition, we combine our method with different architectures. Extensive experiments are conducted to analyze the performance of the PKG-Net with different hyper-parameters and architectures. Finally, in order to verify the effectiveness of our method on Alzheimer's disease diagnosis, we carry out extensive experiments on the ADNI dataset, and compare the results of our method with that of existing methods in terms of accuracy, recall and precision. Furthermore, our network can fully take advantage of the deep 3D convolutional neural network for automatic feature extraction and representation, and thus can avoid the limitation of low processing efficiency caused by the preprocessing procedure in which a specific area needs to be annotated in advance. Finally, we evaluate our proposed framework using two public datasets, ADNI-1 and ADNI-2, and the experimental results show that our proposed framework can achieve superior performance over state-of-the-art approaches. [ABSTRACT FROM AUTHOR]