Enhanced Deep Residual Networks for Single Image Super-Resolution
By Bee Lim, Sanghyun Son, Heewon Kim, Seungjun Nah, Kyoung Mu Lee. 2017
7,355 citations recorded by the source corpus.
Recent research on super-resolution has progressed with the development of deep convolutional neural networks (DCNN). In particular, residual learning techniques exhibit improved performance. In this paper, we develop an enhanced deep super-resolution network (EDSR) with performance exceeding those of current state-of-the-art SR methods. The significant performance improvement of our model is due to optimization by removing unnecessary modules in conventional residual networks. The performance is further improved by expanding the model size while we stabilize the training procedure. We also propose a new multi-scale deep super-resolution system (MDSR) and training method, which can reconstruct high-resolution images of different upscaling factors in a single model. The proposed methods show superior performance over the state-of-the-art methods on benchmark datasets and prove its excellence by winning the NTIRE2017 Super-Resolution Challenge[26].
Subfields: Computer Vision and Pattern Recognition, Media Technology, Efficiency & Compression
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
- Lei Zhang
- Radu Timofte
- Xiaogang Wang
- Trevor Darrell
In Media Technology
In Efficiency & Compression
Citations in the corpus
- The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalization (2020, methodology): “For our experiments, we specifically use the CAE [31] and EDSR [20] architectures as the basis for DeepAugment.”
- Designing a Practical Degradation Model for Deep Blind Image Super-Resolution (2021, methodology): “First, we use a slightly different HR image dataset which includes DIV2K [2], Flick2K [24, 43], WED [30] and 2,000 face images from FFHQ [20] to capture the image prior.”
- The Unreasonable Effectiveness of Deep Features as a Perceptual Metric (2018, neutral): “rain 56.6k 2 CNN-based [7] Train 38.1k 2 Mixed [7] Train 56.6k 2 2AFC–Distort [Trn] – Train 151.4k 2 Traditional [10] Train 4.7k 5 CNN-based [10] Train 4.7k 5 2AFC–Distort [Val] – Val 9.4k 5 Superres [32] Val 10.9k 5 Frame Interp [50] Val 1.9 5 Video Deblur [5] Val 9.4 5 Colorization [47] Val 4.7 5 ”
- The Perception-Distortion Tradeoff (2017, background): “[2] (“primary” variant), EDSR [24], SRResNet variants which optimize MSE”
- ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks (2018, methodology): “We also remove Batch Normalization (BN) [19] layers as in [20] and use residual scaling [21,20] and smaller initialization to facilitate training a very deep network.”
- Image Inpainting for Irregular Holes Using Partial Convolutions (2018, methodology): “We compare with two well-known image super-resolution approaches SRGAN[16] and MDSR+[18] with K =4 in Figure 13.”
- Image Super-Resolution Using Very Deep Residual Channel Attention Networks (2018, methodology): “As investigated in [10, 21], we use only one convolutional layer (Conv) to extract the shallow feature F0 from the LR input”
- Multilevel and Multiscale Network for Single-Image Super-Resolution (2019, methodology): “To make full use of the data, we adopt data augmentation as reference [25] did.”
- Denoising Prior Driven Deep Neural Network for Image Restoration (2018, background): “, SRCNN [22], FSRCNN [31], VDSR [32] and EDSR [42] for image super-resolution, and TNRD [43], DnCNN [24] and MemNet [44] for image denoising.”
- Unsupervised Image Super-Resolution with an Indirect Supervised Path (2019, background): “Further works [16, 17, 19, 26, 35, 29] mainly focus on design of network architecture for performace improvement.”
- Analyzing Perception-Distortion Tradeoff using Enhanced Perceptual Super-resolution Network (2018, methodology): “Our implementation was done in PyTorch and was built on top of the official PyTorch implementation of [31] which was available online.”
- Practical Full Resolution Learned Lossless Image Compression (2018, methodology): “The predictor is based on the super-resolution architecture from EDSR [21], motivated by the fact that our prediction tasked is somewhat related to super-resolution in that both are dense prediction tasks involving spatial upsampling.”
- Learning Convolutional Networks for Content-Weighted Image Compression (2017, methodology): “Recently, deep convolutional networks (CNNs) have achieved unprecedented success in versatile vision tasks [12, 33, 16, 8, 31, 7, 13, 2].”
- Simultaneous Enhancement and Super-Resolution of Underwater Imagery for Improved Visual Perception (2020, background): “Moreover, deep residual networks [42, 30] and residual-in-residual networks [64, 47] are known to be very effective for learning SISR.”
- SwinIR: Image Restoration Using Swin Transformer (2021, methodology): “Most CNN-based methods focus on elaborate architecture designs, such residual learning [43, 51] and dense connections [97, 81], although the performance is improved significantly compared with traditional model-based meth-”
- Deep Back-Projection Networks for Super-Resolution (2018, methodology): “EDSR [30], the winner of NTIRE2017 [43], belongs to this type.”