[Bf-committers] ...depth of field

[Ton Roosendaal]

Hi,

> if I use the convolution method (or a method=A0combined with =20
> =A0convolution) I don't see yet
> how I can do it (convolve) without the full image already generated in =
=20
> R.rectot. Can somebody give me a=A0hint please?


Only at the end of the rendering, a full image is available. Together =20=

with R.rectot you then also have a R.rectz with the zbuffer values.
Doing it internally, the core pixel-rendering loop, is not possible =20
(scanlines, part rendering). It would need a complete new internal =20
structure.
You'll have to decide from the beginning if you want a post-process =20
blur or not...

Suggestion:
What I recognize from animation movies (pixar) is that they make use of =20=

layering a lot. One of the ideas for Blender, was to allow multiple =20
scenes to render together, with each scene having information in it how =20=

it renders over the previous scene. Options you can think of are:
- zbuffer clear (creates 'overlays')
- convolution masks
- blending factors (alpha, mul, add, sub, etc)
- each scene a different 'clip start' and 'clip end' value of course
Just a bit like how 'layers' for photoshop work. And fully animatable =20=

with Ipos!

Personally I think this approach would give artists the most creative =20=

freedom. The quest for 'real' D.o.F. is beyond what Blender offers now, =20=

or what we should strive for.

-Ton-

------------------------------------------------------------------------=20=

--
Ton Roosendaal  Blender Foundation ton@blender.org =20
http://www.blender.org