With a stationary video zoomed in, if watching as a human, more details can be "seen" over multiple frames because we're able to tune out noise and/or combine slightly shifted detail into one coherent understanding of the scene.

Is there any software which can do likewise to extract a frame from video which is higher resolution than the video itself? Perhaps using a traditional algorithm and/or AI.

What is not of interest here is single frame super resolution AI upscaling which only uses one frame as the source.

  • You can probably use the free Hugin panorama stitcher, it is capable of much more than the name suggests. – Orbit Feb 1 at 14:59

You're looking for a concept called 'super-res'. There's tools in python to do it, but most is all one-off and custom due to the amount of tuning needed.

Some cameras are main-streaming it though, so you may see more tools coming out.

You can also look at some astronomical processing software, but you'll have to provide the decoded individual frames. In combination with 'drizzle' you may be able to eek out a better set of resolution without having to go down the programming path.

| improve this answer | |
  • Thanks. Based on your recommendation I found astronomy.stackexchange.com/questions/23179/… is an interesting read on this topic – g491 Feb 3 at 17:54
  • Also several papers utilizing 'AI' based detail infill. I'm not keen on them based upon training datasets and lack of fitment for my need (aka, hand held night photography) – J.Hirsch Feb 3 at 18:59

Another thing you might try (and that is a more comfortable, albeit less controllable) is upscaling the images. There is a lot of free software out there.

| improve this answer | |
  • Thanks for downvoting. Would you care to explain why? – Count Zero Feb 12 at 21:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.