On-demand Learning for Deep Image Restoration

  title={On-demand Learning for Deep Image Restoration},
  author={Ruohan Gao and Kristen Grauman},
  journal={2017 IEEE International Conference on Computer Vision (ICCV)},
  • Ruohan Gao, K. Grauman
  • Published 5 December 2016
  • Computer Science
  • 2017 IEEE International Conference on Computer Vision (ICCV)
While machine learning approaches to image restoration offer great promise, current methods risk training models fixated on performing well only for image corruption of a particular level of difficulty—such as a certain level of noise or blur. First, we examine the weakness of conventional “fixated” models and demonstrate that training general models to handle arbitrary levels of corruption is indeed non-trivial. Then, we propose an on-demand learning algorithm for training image restoration… 

Figures and Tables from this paper

Deep Likelihood Network for Image Restoration with Multiple Degradations
A novel method dubbed deep likelihood network (DL-Net), aiming at generalizing off-the-shelf image restoration networks to succeed over a spectrum of degradation settings while keeping their original learning objectives and core architectures is proposed.
Deep Likelihood Network for Image Restoration With Multiple Degradation Levels
Deep likelihood network (DL-Net), aiming at generalizing off-the-shelf image restoration networks to succeed over a spectrum of degradation levels, is proposed, and extensive experimental results show the effectiveness of the proposed network.
ResDNN: deep residual learning for natural image denoising
A deep convolution neural network with added benefits of residual learning for image denoising is proposed, capable of learning end-to-end mappings from noise distorted images to restored cleaner versions.
X-GANs: Image Reconstruction Made Easy for Extreme Cases
This work presents a new method, called X-GANs, for reconstruction of arbitrary corrupted resource based on a variant of conditional generative adversarial networks (conditional GANs), which has significantly improved the accuracy and quality of image reconstruction.
An Image Restoration Architecture using Abstract Features and Generative Models
An image restoration architecture that enables us to address different types of corruption, regardless of type, amount and location is proposed and has been compared with both traditional approaches like BM3D and with modern deep models like IRCNN and NCSR.
A New Method of Image Restoration Technology Based on WGAN
An image inpainting network based on Wasserstein generative adversarial network (WGAN) distance is proposed, using the Adam algorithm to replace the traditional stochastic gradient descent, and another algorithm to optimize the training used in recent years.
Image Completion Based on Gans with a New Loss Function
This work proposes an image completion method based on the generation adversarial network that can keep the completion regions consistent with the surrounding pixels, which makes the images look more realistic.
Semantic Image Inpainting Vsing Self-Learning Encoder-Decoder and Adversarial Loss
Two new techniques for image inpainting techniques using Deep Convolution Neural Networks (CNN) are proposed and have superior performance in recovering missing regions in deteriorated images over other state-of-art techniques.


Natural Image Denoising with Convolutional Networks
An approach to low-level vision is presented that combines the use of convolutional networks as an image processing architecture and an unsupervised learning procedure that synthesizes training samples from specific noise models to avoid computational difficulties in MRF approaches that arise from probabilistic learning and inference.
Image Super-Resolution Using Deep Convolutional Networks
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
Image Denoising and Inpainting with Deep Neural Networks
A novel approach to low-level vision problems that combines sparse coding and deep networks pre-trained with denoising auto-encoder (DA) is presented and can automatically remove complex patterns like superimposed text from an image, rather than simple patterns like pixels missing at random.
Deep Convolutional Neural Network for Image Deconvolution
This work develops a deep convolutional neural network to capture the characteristics of degradation, establishing the connection between traditional optimization-based schemes and a neural network architecture where a novel, separable structure is introduced as a reliable support for robust deconvolution against artifacts.
Semantic Image Inpainting with Perceptual and Contextual Losses
A novel method for image inpainting based on a Deep Convolutional Generative Adversarial Network that can successfully predict semantic information in the missing region and achieve pixel-level photorealism, which is impossible by almost all existing methods.
Context Encoders: Feature Learning by Inpainting
It is found that a context encoder learns a representation that captures not just appearance but also the semantics of visual structures, and can be used for semantic inpainting tasks, either stand-alone or as initialization for non-parametric methods.
Shrinkage Fields for Effective Image Restoration
  • Uwe Schmidt, S. Roth
  • Computer Science
    2014 IEEE Conference on Computer Vision and Pattern Recognition
  • 2014
This work proposes shrinkage fields, a random field-based architecture that combines the image model and the optimization algorithm in a single unit, and demonstrates state-of-the-art restoration results with high levels of computational efficiency, and significant speedup potential through inherent parallelism.
Perceptual Losses for Real-Time Style Transfer and Super-Resolution
This work considers image transformation problems, and proposes the use of perceptual loss functions for training feed-forward networks for image transformation tasks, and shows results on image style transfer, where aFeed-forward network is trained to solve the optimization problem proposed by Gatys et al. in real-time.
Image Restoration Using Very Deep Convolutional Encoder-Decoder Networks with Symmetric Skip Connections
This paper proposes to symmetrically link convolutional and de-convolutional layers with skip-layer connections, with which the training converges much faster and attains a higher-quality local optimum, making training deep networks easier and achieving restoration performance gains consequently.
A Machine Learning Approach for Non-blind Image Deconvolution
This work relies on a two-step procedure, but learns the second step on a large dataset of natural images, using a neural network, and shows that this approach outperforms the current state-of-the-art on aLarge dataset of artificially blurred images.