| Paper ID | D-1-2.2 |
| Paper Title |
Hallucinating Scenes |
| Authors |
Ting-I Hsieh, Hwann-Tzong Chen, National Tsing Hua University, Taiwan; Chia-Ming Cheng, MediaTek Inc., Taiwan; Yan-Hao Huang, Industrial Technology Research Institute, Taiwan |
| Session |
D-1-2: Machine Learning Techniques for Image & Video |
| Time | Tuesday, 08 December, 15:30 - 17:00 |
| Presentation Time: | Tuesday, 08 December, 15:45 - 16:00 Check your Time Zone |
|
All times are in New Zealand Time (UTC +13) |
| Topic |
Image, Video, and Multimedia (IVM): |
| Abstract |
The goal of this work is to synthesize high-resolution panoramic images of a scene. We propose a system that takes low-resolution panoramic video frames as the input, extracts geometric information as the reference, and produces high-resolution panoramas of the scene as the output. The deep learning module of our system benefits from the structure-from-motion geometry and learns to decouple the content and the style of an input low-resolution patch for hallucinating its high-resolution version. We show that the high-resolution panoramas generated by our system achieve a better quality of detail and color enhancement than those produced by existing super-resolution and style-transfer methods. |