Probability Maps for Noise Functions

I'd like to be able to pass a map to a noise function like cells that would influence the probability of point placement.

To better explain what I'm getting at I'll use some images...

Let's say this is the result of a generic cells noise function...



As you can see in the image, these are the points from which the noise is produced...



What I'd like to do is to apply a mask like so...



... to the points like so...



... so that the result does not generate all the cells you see in the first image.

It would be a "Probability Map" where black would reduce the chance that a point would be plotted in that space.

It would be similar to masking the result like this...



... only the cells wouldn't get clipped.

Nice idea

Thanks. It's something I came up with when trying to find a way to make my Wartorn filter less uniform.

I imagine it would probably come in handy with my SpaceGen filter aswell.

it's unlikely that we'll add probability control to Worley-based noises, but we're considering this for Bombers.

Quote
Richard Bartlett wrote: At present it's "unlikely" which suggests that you don't care for the idea

Richard, my apologies for not being clear. That's definitely not what I meant. English is my second (third?) language, so I'm just not aware of 'hidden meanings' of some commonly used words

So let me correct my answer: It is unlikely that we implement this in the near future, but I don't rule out this possibility for the long term.

The reasons for this are mostly technical. Worley-based noises are the most complex components in Filter Forge, so introducing even minor changes requires lots of testing. Also, since the Worley algorithm can output values in different numerical ranges, depending on the distance function, we normalize the noise output for each of these distance functions, so that the brightness of noise pixels is in the range of 0 to 1, and the energy peak is located at 0.5 (to avoid darkening or lightening when multiple octaves are used). Introducing variable feature point density may require us to rethink our normalization strategy and renormalize everything again, which also requires a significant amount of re-testing. And another reason is that we'll need additional space for the Density parameter in the Settings area (currently 15 parameters only and no scrolling).

Quote
Richard Bartlett wrote: All I'd like to know is if you find the suggestion valuable or not.

Of course it is. Actually, it suggested us a way we can implement this. Early specs for the Worley codebase had a Density input, but we dropped it because it was too difficult to implement -- it required each 'lump' to fade out smoothly, instead of disappearing abruptly, as you're suggesting. With a 'temporally discontinuous' disappearance things would be much easier. However, the normalization problem still remains.

sorry about the timewarp bump

Case closed.