Hi,
Rendering the top left quarter puzzles me:
When I render double sized images, will the frameblur blur before or after
halving the image in post process scaling tab?
As usual, post image's Output scaling happens after post effects. Frameblur
is not an exception..
I mean is there any sense in rendering double sized images?
Frameblur does not need it, but extra resolution naturally improves
antialiasing. In this particular case - simple color channel blurring - the
additional antialiasing could have been included before frameblur simply by
using higher antialiasing level.
The original example of frameblurred GI uses double resolution, because
spreading GI around often causes jagged edges (for example: a shadowed dark
edge of an object can become brightly illuminated). Downscaling after post
effect compensates this kind of reduction of aa quality.
Best regards,
Vesa
BTW: Maybe you already used stochastic aa in your experiment, but anyway,
here's some notes about the issue: I expect that stochastic aa with zero
threshold works better with thin hairs than pure post effect downscaling,
which always uses a regular sampling grid. Anyway, both can be combined to
achieve even higher level of aa.
