Image Super-Resolution Using Very Deep Residual Channel Attention Networks
arXiv:1807.02758
Abstract
Convolutional neural network (CNN) depth is of crucial importance for image super-resolution (SR). However, we observe that deeper networks for image SR are more difficult to train. The low-resolution inputs and features contain abundant low-frequency information, which is treated equally across channels, hence hindering the representational ability of CNNs. To solve these problems, we propose the very deep residual channel attention networks (RCAN). Specifically, we propose a residual in residual (RIR) structure to form very deep network, which consists of several residual groups with long skip connections. Each residual group contains some residual blocks with short skip connections. Meanwhile, RIR allows abundant low-frequency information to be bypassed through multiple skip connections, making the main network focus on learning high-frequency information. Furthermore, we propose a channel attention mechanism to adaptively rescale channel-wise features by considering interdependencies among channels. Extensive experiments show that our RCAN achieves better accuracy and visual improvements against state-of-the-art methods.
arXiv commentTo appear in ECCV 2018
References in corpus (7)
- Sketch-based Manga Retrieval using Manga109 Dataset
- A Learned Representation For Artistic Style
- Residual Dense Network for Image Super-Resolution
- Learnable pooling with Context Gating for video classification
- MemNet: A Persistent Memory Network for Image Restoration
- Learning Deep CNN Denoiser Prior for Image Restoration
- Joint Line Segmentation and Transcription for End-to-End Handwritten Paragraph Recognition
Cited by in corpus (38)
- ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks
- Cross-Scale Internal Graph Neural Network for Image Super-Resolution
- Selective Kernel Networks
- TDAN: Temporally Deformable Alignment Network for Video Super-Resolution
- A Practical Contrastive Learning Framework for Single-Image Super-Resolution
- NTIRE 2024 Challenge on Image Super-Resolution (x4): Methods and Results
- Meta-SR: A Magnification-Arbitrary Network for Super-Resolution
- Gated Multiple Feedback Network for Image Super-Resolution
- Self-Supervised Variational Auto-Encoders
- Deep Fourier-embedded Network for RGB and Thermal Salient Object Detection
- A Single Video Super-Resolution GAN for Multiple Downsampling Operators based on Pseudo-Inverse Image Formation Models
- Learning Digital Camera Pipeline for Extreme Low-Light Imaging
- Searching for Winograd-aware Quantized Networks
- 3DVSR: 3D EPI Volume-based Approach for Angular and Spatial Light field Image Super-resolution
- Video Super-resolution with Temporal Group Attention
- Deep SR-ITM: Joint Learning of Super-Resolution and Inverse Tone-Mapping for 4K UHD HDR Applications
- BAM: A Balanced Attention Mechanism for Single Image Super Resolution
- Super-Resolution Perception for Industrial Sensor Data
- PMC-GANs: Generating Multi-Scale High-Quality Pedestrian with Multimodal Cascaded GANs
- CNN Acceleration by Low-rank Approximation with Quantized Factors
- Difficulty-aware Image Super Resolution via Deep Adaptive Dual-Network
- SamplingAug: On the Importance of Patch Sampling Augmentation for Single Image Super-Resolution
- HDR Denoising and Deblurring by Learning Spatio-temporal Distortion Models
- XCycles Backprojection Acoustic Super-Resolution
- Image Demoireing with Learnable Bandpass Filters
- Pyramidal Dense Attention Networks for Lightweight Image Super-Resolution
- Solving the inverse problem of microscopy deconvolution with a residual Beylkin-Coifman-Rokhlin neural network
- Unsupervised Real Image Super-Resolution via Generative Variational AutoEncoder
- Unsupervised Image Super-Resolution with an Indirect Supervised Path
- NTIRE 2025 Challenge on Image Super-Resolution (x4): Methods and Results
- I3DOL: Incremental 3D Object Learning without Catastrophic Forgetting
- Finding Discriminative Filters for Specific Degradations in Blind Super-Resolution
- SuperMeshing: A New Deep Learning Architecture for Increasing the Mesh Density of Metal Forming Stress Field with Attention Mechanism and Perceptual Features
- Image Super-Resolution via Attention based Back Projection Networks
- Adapting Image Super-Resolution State-of-the-arts and Learning Multi-model Ensemble for Video Super-Resolution
- Viewpoint-Aware Channel-Wise Attentive Network for Vehicle Re-Identification
- Distilling with Residual Network for Single Image Super Resolution
- Careful analysis of XRD patterns with Attention