MSU Super-Resolution Filter
Filter for high-quality video resolution enhancement
Method ideas: Dr. Dmitriy Vatolin
Implementation: Artem Titarenko
This filter increases resolution of video sequence while saving and improving details and reducing artifacts.
Main problems of the current approach:
- Fixed zooming factor (two times). Later it will be generalized.
- No scene change detection.
- Processing speed is considerably higher than for other SuperResolution approaches known to us
- Upscaling factor should be a power of 2. Other scaling factors are also possible, but they result in a relatively lesser quality of video due to use of additional common resampling algorithm
- Different presets with various speed/quality trade-off
- Processing is fully automatic
- Processing can be totally consecutive without look ahead at future frames
- There is a special algorithm for noisy video upscaling which gives smoother images of better visual quality
Since the output resolution of video is very large only parts of sequences are shown.
The first example is from "stefan" sequence:
Topaz Enhance SR
The next example is from "news" sequence:
"news", source video and video processed with MSU Super-Resolution filter
(video is here, 974 KB)
Fragment of the "toy" sequence:
"toy", source video and video processed with MSU Super-Resolution filter
(video is here, 660 KB)
Server size: 8069 files, 1215Mb (Server statistics)
Project sponsored by YUVsoft Corp.
Project supported by MSU Graphics & Media Lab