06-27-2021, 03:46 PM
Thank you very much, PixLab, Ofnuts, rich2005, for your answers 
I've expressed myself badly. I've just verified in a dictionary : "same" might mean "similar". But I meant "exactly identical".
I try to explain:
Sobel:
difference of gaussians (legacy):
Strokes on the darker side can be obtained by firstly "colors->invert"-ing the image. But the result is more noisy (after desaturate, invert, adjust with curves) (I think it's not so easy to denoise without loosing some edges):
GEGL:
legacy:
I've expressed myself badly. I've just verified in a dictionary : "same" might mean "similar". But I meant "exactly identical".
I try to explain:
PixLab Wrote:...snip... between 0.250 and 0.6 ...snip...Yes, that is a little similar. Thanks for the figures (I didn't know which unities the two filters used).
Ofnuts Wrote:...snip... some better ones in Filters -> Edge-detect -> Edge ...snip...
rich2005 Wrote:...snip... Joan Rake gradient norm ...snip... Naggobot smooth sketch ...snip...These filters are great but they stroke the edges astride the edges. This is very good for technical drawings but, for cartoons, it enlarges the eyes of humans and of all the dark zones and this gives a bad result.
Sobel:
difference of gaussians (legacy):
PixLab Wrote:...snip... GEGL ...snip... draw on the outside ...snip... legacy ...snip... draw on the inside ...snip...Yes, one advantage of "difference of gaussians (legacy)" is: it strokes the edges on the darker side of the edge.
Strokes on the darker side can be obtained by firstly "colors->invert"-ing the image. But the result is more noisy (after desaturate, invert, adjust with curves) (I think it's not so easy to denoise without loosing some edges):
GEGL:
legacy:
rich2005 Wrote:check out this from david tschumperleThis is a very interesting aside. Thanks.