Multiple Exemplars-based Hallucinationfor Face Super-resolution and Editing
This addresses face super-resolution and editing for applications like surveillance or photo enhancement, but it is incremental as it builds on prior hallucination methods by incorporating multiple exemplars.
The paper tackles the ill-posed problem of reconstructing high-resolution face images from very low-resolution inputs (e.g., 16x16 pixels) by hallucinating missing details using multiple high-resolution exemplars of the same person, resulting in super-resolved images that are hard to distinguish from real ones in a user study on CelebA.
Given a really low-resolution input image of a face (say 16x16 or 8x8 pixels), the goal of this paper is to reconstruct a high-resolution version thereof. This, by itself, is an ill-posed problem, as the high-frequency information is missing in the low-resolution input and needs to be hallucinated, based on prior knowledge about the image content. Rather than relying on a generic face prior, in this paper, we explore the use of a set of exemplars, i.e. other high-resolution images of the same person. These guide the neural network as we condition the output on them. Multiple exemplars work better than a single one. To combine the information from multiple exemplars effectively, we introduce a pixel-wise weight generation module. Besides standard face super-resolution, our method allows to perform subtle face editing simply by replacing the exemplars with another set with different facial features. A user study is conducted and shows the super-resolved images can hardly be distinguished from real images on the CelebA dataset. A qualitative comparison indicates our model outperforms methods proposed in the literature on the CelebA and WebFace dataset.