> For some definition of "correct". The sampling theorem gives you one > possible definition of what "correct" means. > > And the underlying model can still have infinite resolution. For example a > vector shape, or an insanely high-res raster image. You can manipulate that > without concern, and do anti-aliasing only when you display or rasterize it.
Agreed. > Well, feel free to disagree. It's renamed now, anyway. But to prevent > aliasing was the only reason why this setting was ever introduced. Yes, I understand. It's just that it's the feathering that is trying to work around the aliasing, and not some "aliasing-aware" algorithm. Which is why we had gaps in the brush strokes before. > If you are interested in the dab shape that it produces, then you better > adjust the "hardness" setting. It has exactly the same effect but scales > correctly with the dab size, because it does not depend on the pixel size. Agreed. > That's why I think this feature should be removed as soon as there is a > better method to prevent aliasing. Awaiting your patches :-) > > -- > Martin Renold No, I don't think it should be removed at all. I can see it as a quiet useful feature actually. The only thing that makes me think about it is the fact that we're actually manipulating the hardness to generate the feather effect, which - TBH - I don't know if an artist would be OK with that (given that it's a different - though approximate - result from linearly fading out, which is what - I think - "eathering is about). That said, true AA could be achieved by computing the average intensity value of the function that generates the dab's intensity mask (clamped to 0.0 ... 1.0) given the limits of the pixel's bounds. The only ways I can think of doing this would be integration (relies on a N samples - slow and still only an approximation) and calculating the volume of the intersection of the dab's shape (a half-sphere) and a box (the pixel), but I don't know how to do that, or even if there's a way to do it (without relying on sampling along the way)... -- Micael Dias _______________________________________________ Mypaint-discuss mailing list [email protected] https://mail.gna.org/listinfo/mypaint-discuss
