Accurate Image Super-Resolution Using Very Deep Convolutional Networks

Abstract

We present a highly accurate single-image super-resolution (SR) method. Our method uses a very deep convolutional network inspired by VGG-net used for ImageNet classification \cite{simonyan2015very}. We find increasing our network depth shows a significant improvement in accuracy. Our final model uses 20 weight layers. By cascading small filters many times in a deep network structure, contextual information over large image regions is exploited in an efficient way. With very deep networks, however, convergence speed becomes a critical issue during training. We propose a simple yet effective training procedure. We learn residuals only and use extremely high learning rates ($10^4$ times higher than SRCNN \cite{dong2015image}) enabled by adjustable gradient clipping. Our proposed method performs better than existing methods in accuracy and visual improvements in our results are easily noticeable.

CVPR 2016 Oral

References in corpus (1)

• MatConvNet - Convolutional Neural Networks for MATLAB

Cited by in corpus (23)

• The Perception-Distortion Tradeoff
• Deep Residual Learning for Compressed Sensing CT Reconstruction via Persistent Homology Analysis
• Super Resolution Convolutional Neural Network Models for Enhancing Resolution of Rock Micro-CT Images
• CNN-based Segmentation of Medical Imaging Data
• Compression Artifacts Removal Using Convolutional Neural Networks
• Fast and Accurate Image Super-Resolution with Deep Laplacian Pyramid Networks
• SRPGAN: Perceptual Generative Adversarial Network for Single Image Super Resolution
• Patch-based Progressive 3D Point Set Upsampling
• E-LPIPS: Robust Perceptual Image Similarity via Random Transformation Ensembles
• Framing U-Net via Deep Convolutional Framelets: Application to Sparse-view CT
• Single Image Reflection Removal Using Deep Encoder-Decoder Network
• Image Super-Resolution via Dual-State Recurrent Networks
• On the Relation between Color Image Denoising and Classification
• Deep Stacked Networks with Residual Polishing for Image Inpainting
• AdaIN-Switchable CycleGAN for Efficient Unsupervised Low-Dose CT Denoising
• Deep Residual Network for Joint Demosaicing and Super-Resolution
• Beyond Deep Residual Learning for Image Restoration: Persistent Homology-Guided Manifold Simplification
• Deep Imbalanced Attribute Classification using Visual Attention Aggregation
• Towards WARSHIP: Combining Components of Brain-Inspired Computing of RSH for Image Super Resolution
• Implicit Subspace Prior Learning for Dual-Blind Face Restoration
• Adapting Image Super-Resolution State-of-the-arts and Learning Multi-model Ensemble for Video Super-Resolution
• Deep Interactive Denoiser (DID) for X-Ray Computed Tomography
• Fine-scale Surface Normal Estimation using a Single NIR Image