Super Resolution Debayer?

[Steve Ellington]

This question is mainly targeted to Graeme but anyone feel free to chip it.

I have been reading about the concept of Super Resolution lately which uses temporal information to tease out more resolution from an image. I also just read the great news about RED improving their theoretical resolution limit. My question: Would it be possible to take temporal information into account during the debayer process to further increase the resolution? Or is this something that is happening already?

 

[Graeme Nattress]

I've read an awful lot of super-resolution papers, and it's a very clever technique, that's also very time consuming. I'm not actually too keen on the visual aesthetics of the end results I've seen though.

There's nothing to stop anyone applying such a process in post to any camera system, but I don't think it's a practical replacement for actually capturing the resolution to begin with!

Graeme

 

[Frank Cueto]

The Furnace plug in collection from The Foundry has a plug in that does exactly this.

 

[Steve Ellington]

Thanks for the answer. Would there be any be any resolution benefit at all in doing this during (before?) the debayer? This is purely theoretical because you are 100% right about it being time consuming. The amount of time it would take to process would probably never be worth the small gain in resolution. I tried cleaning up a very noisy Flip Mino shot with the plugin by Topaz -- it did a pretty good job but took two days to render on my 1stGen intel iMac

 

[Matt W.]

This technology is not terribly useful for current cinema cameras. It only works on areas that are both sharp and in motion. But motion blur with a 1/48th second shutter softens moving areas substantially, especially at extremely high resolution. So an extremely small amount of footage (very slow pans and tilts of landscapes, maybe?) would really benefit from this technology. Most of the time, when the camera is moving, motion blur limits resolution more than the mtf of the sensor (or lens).

Your eyes also automatically extrapolate additional resolution from motion (which is why 2kX48fps looks sharper than 2kX24fps). So this makes such technology less necessary and because it's also redundant with what your eyes are doing, you can get weird unpleasant motion artifacts.

The biggest reason no one needs it: you already have 3.2K or whatever after a standard debayer. If that's not sharp enough it will be after deconvolution or usm and especially after adding a little high frequency grain. For DVD upscaling or something this is a cool and useful technology; with red footage its significance is academic at best.