Computational photography

  • 2
  • Idea
  • Updated 1 month ago
A modern high end cellphone can use computational photography, specifically "stacking" mutliple raw photos and combining them together to create an enhanced image with more detail, less noise, and more color information. They can do this in realtime, on the phone itself, it's the only reason the Pixel phones and etc. look so good.

Yet no offline photo program offers this, except for a free open source tool that only works on nvidia GPUs, allows this: www.kandaovr.com/en/newsroom/news/?id=38. All you need are a fast burst of raw photos, something any good, recent camera can provide. Why on earth isn't this available for lightroom and photoshop and etc? Links to Google's own papers, with examples of just what they're able to get out of tiny cellphone sensors, are below.

https://static.googleusercontent.com/media/hdrplusdata.org/en//hdrplus.pdf

https://sites.google.com/view/handheld-super-res/



Photo of Sora Thompson

Sora Thompson

  • 1 Post
  • 0 Reply Likes

Posted 1 month ago

  • 2
Photo of Marlon Bishop

Marlon Bishop

  • 17 Posts
  • 5 Reply Likes
It's novel tech for smartphones to compete with each other based on compact size, and marketing specs, but the final result of what you get on the small sensor cameras using multiple lenses, math or image stacking like Pixel, are not natural looking as you setting for your intended exposure manually on your full frame sensor DSLR. "ISO stacking" introduces all sorts of artifacts, including loss of resolution, for little to no gain on a full frame camera.
 
You also can't get "bursts of the same photo" on a DSLR, without introducing some kind of blur due to throughput and storage latency. Also due to the larger size sensor you are going to have more blur in general unless on a tripod.. Small sensors avoid these issues because the physics of light works out for them (smaller sensor = less handheld motion blur)
I presume Photoshop / Lightroom aims for pros who shoot manually or priority and are getting good enough if not better final results due to the inherent higher resolution and dynamic range, and if not there, are getting better results working in RAW post, where a human using curves to finalize the exposure makes a better picture than an algorithm which can only guess what your intention was at the moment you took the shot.. 

(Edited)
Photo of straannick

straannick

  • 9 Posts
  • 0 Reply Likes
Sora, do you mean something like "Layer/Smart Objects/Stack mode" in Photoshop?
(Edited)