Deep Likelihood Network for Image Restoration with Multiple Degradation Levels
This addresses a practical limitation in image restoration for applications requiring robustness to varying degradation, though it is incremental as it modifies existing networks.
The paper tackles the problem of image restoration networks failing when degradation levels vary, proposing a deep likelihood network (DL-Net) that generalizes off-the-shelf networks to handle multiple degradation levels, with extensive experiments showing effectiveness on tasks like inpainting and super-resolution.
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.