Hacker News Reader

Has anyone tried to stack and super-res the video frames? It is awfully small to see details but it seems like there might be a few clips where a multi-frame-based enhancement could do something.

View on HN

I get the sentiment but please don't make this a thing. You probably already understand that superresolution can only invent a plausible upscaling, no extra information is gained from the pixels - but the general public don't know this. I really think we should be cautious about normalising the use of such tools for conducting forensics on photos.

It sounds like you're referring to machine learning based superresolution? Superresolution by aligning and combining images involves a lot more pixels in total, and so more information.

Can you elaborate on this? If you have multiple frames where the target has moved around and you line up the images couldn’t we get single image with more information?

Yes. But often "multiple" means tens of images and "moved around" does not include video compression where features are moved between frames.

https://m.dpreview.com/articles/0727694641/here-s-how-to-pix...

Reply on news.ycombinator.com