grid cracking: Black spots / sampling errors
Brought to you by:
ltatkinson,
pgregory
Menu
▾
▴
#110 grid cracking: Black spots / sampling errors
Enclosed is a simple test case for the "black spots"
sampling errors that have been reported off-and-on.
The world.rib file contains a single NuPatch arranged
to create a disk, while sample_buster.sl is an imager
shader that does a straight color assignment to any
pixel with an alpha < 1. By rendering geometry with
the constant shader and a contrasting color, any
sampling errors or cracks in the geometry show-up as
isolated pixels - normally, these types of errors would
tend to be masked by filtering and oversampling.
Cheers,
Tim
RIB File
Imager Shader
Logged In: YES
user_id=73804
Additional observations:
Higher pixel sample rates produce more discontinuities,
presumably more samples means more opportunity for error,
although these errors would tend to be hidden better without
the sample_buster shader.
Now that I know what I'm looking for, I've verified that
these discontinuities exist in all primitive types including
conics, polygons, SDS, and NURB patches, although polygons
have noticeably fewer problems than other types.
The distribution of discontinuities is strongly
view-dependent. In a minority of cases I stumbled upon
distributions that appeared to be UV-aligned, although I
haven't been able to reproduce them.
Amusingly, 3delight seems to have this same problem, but
BMRT does not.
Cheers,
Tim
Logged In: YES
user_id=1086736
Originator: NO
As stated in the thread here:
http://www.aqsis.org/xoops/modules/newbb/viewtopic.php?topic_id=1520&forum=2&post_id=8273#forumpost8273
this issue now has a workaround for this particular test case. It's not a proper fix of course, so this bug will remain open.
Logged In: YES
user_id=1086736
Originator: NO
Bug #1525403: "cracks in SubdivisionMesh" is a duplicate of this bug - see there for extra test cases.