Image Restoration by Solving IVP

Recent research on super-resolution (SR) has achieved great success with the aid of deep learning technologies, but, many of them are limited to dealing with arbitrary scaling factors and can only handle fixed scaling factors (e.g., x2, x4). To alleviate this problem, we introduce a new formulation for image super-resolution using an ordinary differential equation parameterized by a convolutional neural network to solve arbitrary scale image superresolution methods. Based on the proposed new SR formulation, we can not only super-resolve images with an arbitrary scale, but also find a new way to analyze the performance of super-resolving process. We demonstrate that the proposed method can generate high-quality images with arbitrary scaling factors, unlike conventional SR methods.