Image Super-Resolution Using Deep Convolutional Networks
By Chao Dong, Chen Change Loy, Kaiming He, Xiaoou Tang. 2014
9,416 citations recorded by the source corpus.
We propose a deep learning method for single image super-resolution (SR). Our method directly learns an end-to-end mapping between the low/high-resolution images. The mapping is represented as a deep convolutional neural network (CNN) that takes the low-resolution image as the input and outputs the high-resolution one. We further show that traditional sparse-coding-based SR methods can also be viewed as a deep convolutional network. But unlike traditional methods that handle each component separately, our method jointly optimizes all layers. Our deep CNN has a lightweight structure, yet demonstrates state-of-the-art restoration quality, and achieves fast speed for practical on-line usage. We explore different network structures and parameter settings to achieve trade-offs between performance and speed. Moreover, we extend our network to cope with three color channels simultaneously, and show better overall reconstruction quality.
Subfields: Computer Vision and Pattern Recognition, Media Technology
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Ming-Hsuan Yang
- Dacheng Tao
- Chen Change Loy
- Xiaogang Wang
- Trevor Darrell
- Lei Zhang
- Jitendra Malik
In Media Technology
Citations in the corpus
- Image Super-Resolution via Iterative Refinement (2021, neutral): “sion based and trained with an MSE loss [49], [52], [54], [55], [56].”
- Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network (2016, methodology): “We validate the proposed approaches using images from publicly available benchmark datasets and compare our performance against state-of-the-art works including SRCNN [8], SelfExSR [24] and DRCN [27].”
- Perceptual Losses for Real-Time Style Transfer and Super-Resolution (2016, methodology): “Recently [1] achieved excellent performance on single-image super-resolution using a three-layer convolutional neural network with a per-pixel Euclidean loss.”
- Unsupervised Cross-Domain Image Generation (2016, methodology): “Network g maps f ’s 256D representations to 64 × 64 RGB images through a network with 5 blocks, each consisting of an upscaling convolution, In order to upscale the 64 × 64 output to print quality, we used the method of Dong et al. (2015), which was shown to work well on art.”
- Amortised MAP Inference for Image Super-resolution (2016, methodology): “Most current single image SR methods use empirical risk minimisation, often with a pixel-wise mean squared error (MSE) loss (Dong et al., 2016; Shi et al., 2016).”
- UberNet: Training a Universal Convolutional Neural Network for Low-, Mid-, and High-Level Vision Using Diverse Datasets and Limited Memory (2016, background): “While only considering works that apply to a single, static image we can indicatively list successes of CNNs in superresolution [24], colorization [55], boundary detection [6,28,49,107], symmetry detection [91], interest point detection [109], image descriptors [35, 92, 110], surface Input…”
- Recent advances in convolutional neural networks (2015, methodology): “ut. Recently, deconvolution has been widely used for visualization [8, 35{37], recognition [38{41], localization [42], semantic segmentation [34], visual question answering [43], and super-resolution [44, 45]. 3.1.3. Dilated Convolution Dilated CNN [46] is a recent development of CNN that introduces”
- Hierarchical Neural Architecture Search for Single Image Super-Resolution (2020, background): “These two kinds of blocks/layers often follow a two-level architecture, where the network-level architecture determines the positions of the upsampling layers ( e.g. , SRCNN [7] and LapSRN [8]) and the cell-level architecture controls the computation of each block/layer ( e.g. , RCAB [9]).”
- Multi-View Silhouette and Depth Decomposition for High Resolution 3D Object Representation (2018, background): “Modern approaches to super-resolution are built on deep convolutional networks [5, 45, 27] as well as generative adversarial networks [18, 13] which use an adversarial loss to imagine high-resolution details in RGB images.”
- Real-Time Speech Frequency Bandwidth Extension (2020, methodology): “Inspired by the early success in image super-resolution [4], end-to-end audio-based solutions were proposed, based on wave-to-wave UNet [5], WaveNet [6, 7], hybrid time/frequency domain models [8].”
- Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network (2016, methodology): “Secondly, interpolation methods typically used to accomplish the task, such as bicubic interpolation [7, 44, 3], do not bring additional information to solve the ill-posed reconstruction problem.”
- Combining Markov Random Fields and Convolutional Neural Networks for Image Synthesis (2016, background): “While an inverse process can be defined [4, 5, 8, 9, 10, 20, 28, 29], it reamins difficult to control: For example, simply maximizing the class-excitation of the network leads to hallucinatory patterns [21].”
- Accurate Image Super-Resolution Using Very Deep Convolutional Networks (2015, methodology): “Our network is very deep (20 vs. 3 [6]) and information used for reconstruction (receptive field) is much larger ( 41 × 41 vs. 13 × 13 ).”
- Checkmate: Breaking the Memory Wall with Optimal Tensor Rematerialization (2019, background): “Deep learning is rapidly pushing the limits of memory capacity on neural network accelerators as researchers train neural networks on high-resolution images [15, 32, 55], 3D point-clouds [10, 59], and long Natural Language Processing (NLP) sequence data [11, 14, 56].”
- Deeply-Recursive Convolutional Network for Image Super-Resolution (2015, methodology): “The network takes an interpolated input image (to the desired size) as input x and predicts the target image y as in SRCNN [5].”
- Deep Laplacian Pyramid Networks for Fast and Accurate Super-Resolution (2017, methodology): “Experimental results demonstrate that our method is faster than several CNN based super-resolution models, e.g., SRCNN [7], SCN [33], VDSR [17], and DRCN [18].”
- Image Super-Resolution Using Very Deep Residual Channel Attention Networks (2018, methodology): “Manga109 (4×): YumeiroCooking HR Bicubic SRCNN [1] FSRCNN [2] VDSR [4] PSNR/SSIM 24.”
- Pre-Trained Image Processing Transformer (2020, background): “For the super-resolution, Dong et al . propose SRCNN [20, 21] which are considered as pio-neering works introducing end-to-end models that reconstructs HR images from their LR counterparts.”