OpenEXR 2.0 (deep image compositing)

General Forums Blender and CG Discussions
1

i today found out about deep image compositing. it will be in OpenEXR 2.0. this is groundbreaking.

http://www.deepimg.com/
http://www.deathfall.com/forums/content.php?7135-Deep-Compositing-will-be-supported-in-OpenEXR-2.0-and-Nuke
http://www.sfdm.scad.edu/faculty/mkesson/vsfx502/wip/fall11/kevin_george/deep/index.html

this will be possible in blender right? :evilgrin:

so what do you think?

1 Like
2

Blender can render to openexr
Check the render formats its there

3

Looks very cool! Interesting to see how this will evolve.

4

Render to OpenEXR 1.0 perhaps, but maybe not OpenEXR 2.0.

5

http://wiki.blender.org/index.php/Doc:2.4/Manual/Composite_Nodes/Types/Color and search for “Z-Combine Node”. Guess there is a way to simulate multiple values per pixel stuff (like Z-transparency).

Still the idea is nice. But guess this 2.0 thing gives it a bit of a spam-marketing stench.

6

Apparently the newer standards allow for working with images that have partial transparency which apparently (I haven’t tested) can cause issues with Z-depth / deep compositing.

7

is this something similar to light field cameras ?

8

@cekuhnen no, the light field cameras are z-depth based depth of field, which we can already do using the defocus node in the compositor

Deep image compositing is useful for say compositing two different layers together where there are alot of alphas going on.

Right now, z-depth doesnt honor any sort of alpha mask nor alpha value… say you have a grass field, each blade of grass would be slightly transparent… Say you use a z-combine the grass with say a box, each rendered seperatly, the blades of grass would be totally solid… If Deep image compositing was enabled, you would get the true alpha value of that blade of grass and you would be able to get that fade off.

9

You’re also able to occlude objects using the Z-depth pass, which is neat.

I think that’s what the recent Autodesk Maya viewport 2.0 was showing off; A cube was dragged over a 2D plane and occluded by the depth values of the image.

Apparently the images are typically in the hundreds of MB’s to GB’s per frame though!!