Hello and thank you for the link!
Video Enhancer tries to be a practical tool, so we have to use simple and effective methods, unlike scientific papers like the one you mention.
I've just run their program on a 320x180 video of 80 frames and upsized it 4x to 1280x720.
On my Intel Core2 Quad (4 cores) their program worked 17 minutes (13 seconds for one frame) while Video Enhancer processed the same file in just 8 seconds (0.1 second for one frame), so it's 130 times faster in maximum quality mode. It's funny how they call their method "simple and efficient" and "very fast".
And here is a sample from results:
http://stuff.infognition.com/smiths13shan.jpghttp://stuff.infognition.com/smiths13ve196.jpgOpen these two images in two tabs of your browser and switch back and forward to compare.
I don't think anyone can tell Shan's result is better than VE. It's got more contrast, but that's really a distortion, not an improvement (colors are wrong, see original:
http://stuff.infognition.com/smiths13.jpg ). VE result has more details and is less jagged.
In their paper and video they used a very old version of VE (released in 2006) for comparison, it's got much better since then.
As for blur, VE does not do any deconvolution or deblurring. In one step of SR it upsizes video 2x in both dimensions, so each original pixel turns into 4 new pixels. VE considers input image as good signal and makes sure it does not bring distortions, so if you average these 4 pixels you get back just the original one.
