12-16-2019, 06:01 AM
(12-15-2019, 11:46 PM)Ofnuts Wrote: Not very clear what the perceived problem is.
AFAIK Gimp 2.10 always work in 32bit-FP internally which is all for the better, in 8-bit round-off errors creep in very quickly: do a bump map in 2.8 and compare with 2.10 and you'll see there is no real debate. Making the image 32-bit FP makes processing faster (no conversions) at the cost of 4x the RAM.
Export artifacts may be be due to the export format and/or the format parameters (compression, etc...).
So perhaps the issue is that the 32-floating is making it appear like the image has enough data to make my transformations, but when it tries to apply them to the 8-bit source they can't be reproduced?
I doubt the artifacts are from the export settings I selected. I get this result when exporting to PNG at level 9 (both with and without gamma embedded), and when exporting to uncompressed TIF. If there's a more lossless option than those I should be trying, let me know.
Do you not see the problem on the image, or do you mean you don't know what could be causing it? The problem is the export artifacts, I just don't know how to address them because I don't know what's causing them.