Focus distance as metadata

[Jack Kelly]

This is old news but these Stanford academics are doing some remarkably interesting stuff with light field cameras:

http://graphics.stanford.edu/papers/lfcamera/

Photographs digitally refocused at different depths, computed after a single exposure of our light field camera. The fourth image shows what a conventional camera would have produced.

Imagine if the focus distance was just another bit of non-destructive metadata recorded by the camera. In post, sharpening an out-of-focus image would be as simple as shifting the focus distance metadata.

[Gunleik Groven]

This is old news but these Stanford academics are doing some remarkably interesting stuff with light field cameras:

Imagine if the focus distance was just another bit of non-destructive metadata recorded by the camera. In post, sharpening an out-of-focus image would be as simple as shifting the focus distance metadata.

Yup. That's cool. But will demand a high F-stop/deep/DOF to sorta work, if you're not into re-inventing optics...

And: How much do you really want to push into post?

But fascinating software, for sure. I can see the use of this for VFX shots. Definitely!

[Erik Widding]

These cameras throw away a lot of spacial information (pixels) to get the depth encoded in a way that can be monkeyed with later. Very very cool tech, but there is a trade off. Fortunately, it isn't a significant trade off in f/number of the system. Effectively, they get a 2D picture at a bunch of different depths, and later process for desired effect.

So (having not read the paper in awhile, my math may be off) you might get S35 resolution from a 645 sized sensor.

[Joseph Ward]

Why not have a high DOF then focus in post?

[Felix K.]

Yup. That's cool. But will demand a high F-stop/deep/DOF to sorta work, if you're not into re-inventing optics...

That is not true, or maybe, they are re-inventing optics.

Read the paper on their page...