Deep Likelihood Network for Image Restoration with Multiple Degradation Levels
arXiv:1904.09105
Abstract
Convolutional neural networks have been proven effective in a variety of image restoration tasks. Most state-of-the-art solutions, however, are trained using images with a single particular degradation level, and their performance deteriorates drastically when applied to other degradation settings. In this paper, we propose deep likelihood network (DL-Net), aiming at generalizing off-the-shelf image restoration networks to succeed over a spectrum of degradation levels. We slightly modify an off-the-shelf network by appending a simple recursive module, which is derived from a fidelity term, for disentangling the computation for multiple degradation levels. Extensive experimental results on image inpainting, interpolation, and super-resolution show the effectiveness of our DL-Net.
Accepted by IEEE Transactions on Image Processing; 13 pages, 6 figures
References in corpus (10)
- Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
- Understanding deep learning requires rethinking generalization
- The Perception-Distortion Tradeoff
- Deep Image Prior
- EdgeConnect: Generative Image Inpainting with Adversarial Edge Learning
- Amortised MAP Inference for Image Super-resolution
- Differentiable Convex Optimization Layers
- Learning a Single Convolutional Super-Resolution Network for Multiple Degradations
- Physics-Based Generative Adversarial Models for Image Restoration and Beyond
- Texture Transform Attention for Realistic Image Inpainting