My suggestion is linked to Photoshop usage in still-image 3D composition. Specifically, it is about new layer blending mode what makes usage of depth channel (Z-buffer) information provided by rendering software.
Very often I would like to be capable of color correction of a single particular object of a 3D rendering. Though it is possible to make appropriate color correction in 3D package - this is nasty as that require additional rendering of the complete scene what is normally pretty time consuming.
Ideally, I would like to do such minor color corrections is a compositing software - in case of stills - in Adobe Photoshop.
But... how to do this?
I tried to make two render passes:
A) rendering the scene having one particular object set invisible to camera (but visible to GIobal Illumination solution)
B) rendering the scene having all objects but this one set invisible to camera (but visible to GIobal Illumination solution)
Then - these two images can be composited in Photoshop as separate layers.
That could work but it fails in case given object intersects with other in 3D space.
Normaly, 3D renderer only renders these object's parts what are visible to 3D camera, so these parts what are obscured by other object are not renedered. But if I render given object "separately" - then I have all its visible parts being rendered - even those what are normally invisible as obscured by intersecting object. And in his case alpha channel provides unsufficient information to perform appropriate selection to remove unneeded parts of the image.
Normally, this information if provided with depth channel: the Z-buffer what can be rendered in an extra pass. This is a grayscale image where shades of gray provide information about given pixel's distance to camera. But unfortunately, I can't see any use of it in Photoshop composition.
Ideally, some specialized "layer blending mode" should be implemented in Photoshop. It could work in the following manner:
Presuming there are two Photoshop layers:
- bottom layer containing (A) render image masked with its Z-buffer channel
- top layer containing (B) render image masked with its Z-buffer channel
Using new blending mode, only these pixels would be revealed what are actually "closer" based on appropriate Z-buffer data comparison between (A) and (B).
Is this suggestion worth implementation?
IMHO that would significantly faciliated 3D composition in Photoshop.