The Persistence of Vision Raytracer (POV-Ray).
This is the legacy Bug Tracking System for the POV-Ray project. Bugs listed here are being migrated to our github issue tracker. Please refer to that for new reports or updates to existing ones on this system.
Opened by Jaime Vives Piqueres - 2011-01-28
Last edited by Christoph Lipka - 2012-06-21
FS#191 - Using interpolated image_maps in functions results in pixel-sized dot-artifacts
Using interpolated image_maps in functions results in pixel-sized dot-artifacts when using the functions back into pigments.
This problem doesn’t shows using the same code on POV-Ray 3.6.
I qualified it as “low severity” because is not going to happen to most users: it will show only when using some advances techniques, for example when you want to decompose an image_map into the RGB components, perform operations, and mixing them back with an averaged pigment (example attached).
interpolation-in-functions-bu...
(55.3 KiB)
interpolation-in-functions-bu...
(1.5 KiB)
interpolation-in-functions-bu...
(48.1 KiB)
interpolation-in-functions-bu...
(0.5 KiB)
Thursday, 21 June 2012, 22:58 GMT
Reason for closing: Fixed
Additional comments about closing:
no news, so I guess the fix was
effective
I tested it with 3.7RC3 on linux: the issue is more visible without AA.
Could it be some clipping of the function value done differently (with round error in the process) ?
For instance, a function might value in theorie between 0 and 1.0; but it happens in fact (due to artefact of numerical calculus, so not for a pure mathematician), than it evolves between -0.000001 and 1.0001 (I have seen such result, such as computing the dot product of two unit vectors... it should stay below (or equal to) 1, but requesting the arcos of it sometime generate error (it was bigger than 1.)
Could the artefact be due to the rollover of the value ? (like with modulo (%))
This is indeed what is happening; it’s a precision problem in the interpolation code.
Looks like I introduced the issue with change #4962; color computation in Interp() was done with double precision before, so I'll re-establish that.
Done. See change #5400.
Tested: Artefact gone. Seems solved with #5400.