Deep Likelihood Network for Image Restoration With Multiple Degradation Levels

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 single particular degradation level, and their performance deteriorates drastically when applied to other settings. In this paper, we propose deep likelihood network (DL-Net), aiming at generalizing off-the-shelf succeed over spectrum levels. We slightly modify an by appending simple recursive module, which is derived from fidelity term, for disentangling the computation multiple Extensive experimental results on inpainting, interpolation, super-resolution show effectiveness our DL-Net.