Single Image Super-Resolution Using Multi-scale Convolutional Neural Network

Jia, Xiaoyi; Xu, Xiangmin; Cai, Bolun; Guo, Kailing

doi:10.1007/978-3-319-77380-3_15

Cited by 10 publications

(9 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There is a good survey on DL approaches for image processing and computer vision related tasks, including image classification, segmentation, and detection [102]. For examples, single image super-resolution using CNN method [103], image denoising using block-matching CNN [104], photo aesthetic assessment using A-Lamp (Adaptive Layout-Aware Multi-Patch Deep CNN) [105], DCNN for hyperspectral imaging segmentation [106], image registration [107], fast artistic style transfer [108], image background segmentation using DCNN [109], handwritten character recognition [110], optical image classification [111], crop mapping using high-resolution satellite imagery [112], object recognition with cellular simultaneous recurrent networks and CNN [113]. The DL approaches are massively applied to human activity recognition tasks and achieved state-of-the-art performance compared to exiting approaches [114][115][116][117][118][119].…”

Section: Image Processing and Computer Visionmentioning

confidence: 99%

A State-of-the-Art Survey on Deep Learning Theory and Architectures

et al. 2019

View full text Add to dashboard Cite

In recent years, deep learning has garnered tremendous success in a variety of application domains. This new field of machine learning has been growing rapidly and has been applied to most traditional application domains, as well as some new areas that present more opportunities. Different methods have been proposed based on different categories of learning, including supervised, semi-supervised, and un-supervised learning. Experimental results show state-of-the-art performance using deep learning when compared to traditional machine learning approaches in the fields of image processing, computer vision, speech recognition, machine translation, art, medical imaging, medical information processing, robotics and control, bioinformatics, natural language processing, cybersecurity, and many others. This survey presents a brief survey on the advances that have occurred in the area of Deep Learning (DL), starting with the Deep Neural Network (DNN). The survey goes on to cover Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), including Long Short-Term Memory (LSTM) and Gated Recurrent Units (GRU), Auto-Encoder (AE), Deep Belief Network (DBN), Generative Adversarial Network (GAN), and Deep Reinforcement Learning (DRL). Additionally, we have discussed recent developments, such as advanced variant DL techniques based on these DL approaches. This work considers most of the papers published after 2012 from when the history of deep learning began. Furthermore, DL approaches that have been explored and evaluated in different application domains are also included in this survey. We also included recently developed frameworks, SDKs, and benchmark datasets that are used for implementing and evaluating deep learning approaches. There are some surveys that have been published on DL using neural networks and a survey on Reinforcement Learning (RL). However, those papers have not discussed individual advanced techniques for training large-scale deep learning models and the recently developed method of generative models.

show abstract

Section: Image Processing and Computer Visionmentioning

confidence: 99%

A State-of-the-Art Survey on Deep Learning Theory and Architectures

et al. 2019

View full text Add to dashboard Cite

show abstract

“…It has been found to be highly effective and is also the most commonly used approach in diverse computer vision applications (Guo et al 2016). In addition, a large number of studies have investigated the CNN model for image processing, including image denoising (Jain and Seung 2008;Xuejiao et al 2015;Zhang et al 2017), image restoration (Cheong and Park 2017;Deepak and Ghanekar 2017;Dong et al 2016;Jia et al 2017;Liu et al 2016;Samuel et al 2015;Yamanaka et al 2017) and image segmentation (Liu et al 2015;Long et al 2015). In this paper, a convolutional neural network reconstruction (CNNR) method that combines super-resolution and image segmentation is used to generate high-resolution segmented images based on low-resolution tomographic µ-CT images and high-resolution two-dimensional sections such as SEM image(s).…”

Section: Introductionmentioning

confidence: 99%

Porous Structure Reconstruction Using Convolutional Neural Networks

et al. 2018

View full text Add to dashboard Cite

The three-dimensional high-resolution imaging of rock samples is the basis for pore-scale characterization of reservoirs. Micro X-ray computed tomography (µ-CT) is considered the most direct means of obtaining the three-dimensional inner structure of porous media without deconstruction. The micrometer resolution of µ-CT, however, limits its application in the detection of small structures such as nanochannels, which are critical for fluid transportation. An effective strategy for solving this problem is applying numerical reconstruction methods to improve the resolution of the µ-CT images. In this paper, a convolutional neural network reconstruction method is introduced to reconstruct high-resolution porous structures based on low-resolution µ-CT images and high-resolution scanning electron microscope (SEM) images. The proposed method involves four steps. First, a three-dimensional low-resolution tomographic image of a rock sample is obtained by µ-CT scanning. Next, one or more sections in the rock sample are selected for scanning by SEM to obtain high-resolution two-dimensional images. The high-resolution segmented SEM images and their corresponding low-resolution µ-CT slices are then applied to train a convolutional neural network (CNN) model. Finally, the trained CNN model is used to reconstruct the entire low-resolution three-dimensional µ-CT image. Because the SEM images are segmented and have a higher resolution than the µ-CT image, this algorithm integrates the super-resolution and segmentation processes. The input data are low-resolution µ-CT images, and the output data are high-resolution segmented porous structures. The Yuzhu Wang

show abstract

“…The RVL-UTA team presents a novel network for thermal image super-resolution (SR) called the Multiscale Residual Channel Attention Network (MSRCAN). The architecture is inspired by state-of-the-art methods to recover details from low-resolution (LR) RGB images such as: very deep residual channel attention networks (RCAN) [13], learning a mixture of deep networks for single image SR (MSCN) [7], and multiscale convolutional neural networks (CNNs) (MSSR) [3]. RCAN allows deeper CNN models, which result in more feature representation.…”

Section: Rvl-utamentioning

confidence: 99%

Thermal Image Super-Resolution Challenge - PBVS 2021

Rivadeneira

Sappa

Vintimilla

et al. 2021

2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)

View full text Add to dashboard Cite

This paper summarizes the top contributions to the first challenge on thermal image super-resolution (TISR), which was organized as part of the Perception Beyond the Visible Spectrum (PBVS) 2020 workshop. In this challenge, a novel thermal image dataset is considered together with stateof-the-art approaches evaluated under a common framework. The dataset used in the challenge consists of 1021 thermal images, obtained from three distinct thermal cameras at different resolutions (low-resolution, mid-resolution, and high-resolution), resulting in a total of 3063 thermal images. From each resolution, 951 images are used for training and 50 for testing while the 20 remaining images are used for two proposed evaluations. The first evaluation consists of downsampling the low-resolution, midresolution, and high-resolution thermal images by ×2, ×3 and ×4 respectively, and comparing their super-resolution results with the corresponding ground truth images. The second evaluation is comprised of obtaining the ×2 superresolution from a given mid-resolution thermal image and comparing it with the corresponding semi-registered highresolution thermal image. Out of 51 registered participants, 6 teams reached the final validation phase.

show abstract

Single Image Super-Resolution Using Multi-scale Convolutional Neural Network

Cited by 10 publications

References 26 publications

A State-of-the-Art Survey on Deep Learning Theory and Architectures

A State-of-the-Art Survey on Deep Learning Theory and Architectures

Porous Structure Reconstruction Using Convolutional Neural Networks

Thermal Image Super-Resolution Challenge - PBVS 2021

Contact Info

Product

Resources

About