On Feb 22, 10:56 pm, finecur wrote:
Hi, anyone know how the "angled strokes" filter was implemented? I am interested in the mathematic algorithm.
Thank you very much.
ff
Hi ff
You ask for a mathematical answer, that’s too abstract for me. But I can tell you roughly how it could be done procedurally.
The first step would be to define the size and falloff profile of the stroke areas.
This could be done with any noise, like cells or stones, but I would probably use Perlin because it’s fast – ultimately this would be a stylistic decision and could be broke out for user control. Any number of stroke areas could be defined by separating the image according to things like saturation, hue, value or even frequency.
Next would be to define the stroke pattern and angle.
This would be easy, just stretch some Perlin and scale it way down. It would basically look like scratches. The angle could be taken care of by placing other layers of stretched Perlin out of phase, and in separate regions, and rotating them – something else that the user could have access to.
Then the stroke lift and distribution would need to be defined. I would do this with cells noise because the value gradients are smooth but again that would be a stylistic decision.
Next would be to combine the stroke regions, pattern, angle, lift and distribution.
This could easily be done by using multiply/darken screen/lighten blending.
This layer would then be used with a very small amount of refraction to effect the shape of the image.
That’s the easy part. The next steps, and the ones that make it pretty, is to use the various layers that had been defined earlier, or even some completely new ones, to abstract color information from the image and reapply it in a way that achieves the look you’re going for. I sure this is not the kind of answer you are looking for but maybe it will give you some direction or at least some search terms to look for.
Good luck,
Ron
