UI3
To our knowledge (and NVIDIA's for that matter), no other SR technology is capable of achieving that high an order of up-resolving without introducing hallucinated (synthetic) pixels. {Making it worthless for computer vision and indefensible as forensic evidence.} The main reason for this is because a camera's imaging pipeline is built for human eye consumption and based on a 42yo colorization patent that radically reduces the resolvability of video at output so we humans with our weak eyesight can enjoy prettier colorized pictures.
Our patented pipeline bypasses that 42 year old limitation (which is why our cameras are dual sensor) and uses deep learning to learn and then revert the native degradation model of any particular imaging hardware pipeline. It just requires a massive amount of computational power.
Five years ago it was not viable because of the limits of computational cycles per dollar. But four years ago NVIDIA introduced a GPU capable of 7.5 Tflops for $999 MSRP. Since that time they have 10x'd that compute power per dollar on their latest generations of chips. Essentially giving everyone access to supercomputing. We saw that coming early and have been spending a few years building a patented imaging pipeline designed to take advantage of such an exponential increase in available compute cycles per dollar. All that was needed was a 'supercomputer' in the pipeline. Something that would have costs a half a million bucks 10 years ago but today can be bought online for a couple grand.
Enough about our thing though. The best solution I can recommend to the original poster looking to clean up video shot with cameras designed for weak human eyesight is VideoCleaner. Doug Carner is the guy you want to talk to about it.