Video Upscaling with Spatio-Temporal Self-Similar Examples
International Conference on Pattern Recognition (ICPR)
We propose a new example-based video upscaling technique that exploits self-similarity among patches of a video in both space and time. We encode image patches with over-complete dictionaries constructed in a local spatio-temporal neighborhood, and establish temporal correspondence using modern optical flow techniques. The resulting method performs favorably compared to the state-of-the-art in super-resolution techniques.