Editing Noise Into Terrain - the Burpwallow Way

[su_liam]

Your example seems to resolve itself into a black and white flat image.

If I understand you correctly. That's what it's supposed to do. At this stage, it is just a grayscale heightfield without texturing or even faux relief. Hopefully the second part of my tut will clarify one way of making it look like a terrain. Another way would be to plug it into a 3d renderer such as Bryce.

Do you ever vary the levels you uploaded to use the heights, depths, and gradients from your initial cloud layer?
.

I'm not sure I understand what you're asking.

One thing I am looking into is basing my texture on a slightly different noise. For example, take the original HF, reduce the range with levels, perhaps adding a bias towards low or high values, add a gradient from light gray in the north to dark gray in the south(or viceversa in the southern hemisphere), then add in an independent cloud noise. Base your texturing gradient maps on that noise source.

Is this the sort of thing you are getting at?

[su_liam]

Maybe you can add a JPG to your first post, so we at least see a thumbnail and what this is all about? Might interest more people to try it out also

Good point. Maybe tomorrow...
<No smiley for tired out of my mind>

[ravells]

Fab stuff, Su Liam - can't wait to try this out when I get home!

[su_liam]

Here's a little something I put together this morning to practice the technique. It's looking pretty powerful. Still no climate coloring, but I'll be working on that.

[ravells]

Here's my go. I find it really tough to get just the right shade of gray for the heightmap when trying to raise and lower terrian while keeping the finer detail...got to keep trying and experimenting, it's a really cool method.

The only thing is that there is not really a lot of control over where the mountains go (unless I've misunderstood completely which is quite likely) as you're reliant on the initial cloud map to a great extent

Thanks Su Liam!

[su_liam]

Hey, Ravells, good start. The first problem I see is my fault. You, in an unfortunate act of English comprehension, applied the Sea Mask as a mask to the sea. Actually, I should have named it the Land Mask. Threshold selects areas higher than the threshold as white. If you are using NW lighting, which it looks like you are, that places the sea areas as higher than the surrounding land. So first invert your masks. I think that will help some.

As to keeping finer detail. The key to this, paradoxically, is a large, soft brush. Fine detail in editing obliterates detail from noise. In fact, in my latest effort I applied a Gaussian Blur to each of the editing layers.

If an area seems to have a lot of relief where you desire flatness, try painting in a lighter shade of gray into the Pin Light layer. OR... In a two step process, try darkening the area slightly in the Multiply layer then raise the area back up with the Linear Light layer. Make sure the Linear Light editing is very soft and blurred. The first option is pretty easy once you get a feel for it, but the second is a lot more nuanced and controllable. I, personally, have yet to make the multiply and add-back method work reliably, but I have very high hopes for it.

[ravells]

Thanks Su Liam! I actually did invert the masks but for the moment, until I work out how to get good results with the land layer am ignoring the sea layer for now (which is after all the same method). I ended up using a bit of gaussian blur too, but as I said, the real trick to this (I think) is finding the right gray setting so it makes a enough of a difference, but not so much that it overwhelms the underlying detail.

Btw, I imported the heightfield into Bryce to see what the result would be but the relative elevations were huge and it needed some smoothing but nothing insurmountable. I'm thinking what might be cool is to make multiple renders in Bryce of different textures (forest, desert etc) and use layer masks in PS to hide / show them. Saves having to use a displacement filter for each of them too.

Edit: also one of the steps I didn't use was the levels adjustment layer - I couldn't find the controls in it you mentioned. I'm using CS2 - is it a post CS2 thing?

[su_liam]

The correct shade of gray is a real judgement call. It helps, because of the way blending and transparency works, to fill in Multiply layers with white and ** Light layers with 50% gray(RGB12. For Multiply, I usually paint(Normal blend mode) about 78% gray to start with. I then set the paintbrush to Multiply mode and 85-95% gray, or Linear Light mode at about 100-125RGB (to lower) or 130-155RGB (to raise) in order to make adjustments. With the Pin Light layer your guess is as good as mine (or better). Mostly, I set a value, paint it in, cuss in fluent Anglo-Saxon, hit Undo and try again.

Pin Light is horribly confusing. It apparently acts as a clamp. If the value is much less than 128, it clamps high points down. If the value is much greater than 128, it clamps low points up. Exactly what it clamps to is a mystery to me. If you can figure out the algorithm, I'd be much in debt if you'd let me know.

Using 16-bit would help quite a bit with the huge terracing artifacts in Bryce, but the Clouds filter doesn't work in 16-bit. You'll have to supply your own noise in that case. And..., you'll have to go back to 8-bit for texturing work and Gradient mapping. No biggie, but it does make things slower.

I've been experimenting with global maps using spherical noise imported from planetGenesis. Not with much success, so far. I can't seem to manage Pin Light...

I'm using Photoshop CS, Old School. If I had $300 to spare, I'd be down taking advantage of my student discount to snag CS4!

In the images, I've tried to highlight important buttons and settings. The first image shows where the options button is on the Levels dialog. The second image highlights important parts of the Options dialog: algo selection, lower clip and upper clip. In the event that the Options dialog has been lost in later versions(?!?), the third image might give a sense of how to manually set up levels to map the output to a narrower histogram. Hopefully this will help. Shouldn't be hard to find a better explanation on the intertubes somewhere.

[Sigurd]

For those who want to pull apart the whole file that Su is generously sharing, here is the whole thing shrunk to 10% size to make it small. It should have all the layer, layer order, and design decisions intact.


I actually am opening this file beside my own experiments to compare notes.


Please, nobody complain about the image view at 10%. This is not anyone's idea of an acceptable dimension for display. At 361k its actually smaller than screenshots for all the settings.

This is exported from CS3.

[su_liam]

Thanks, Sigurd.
Here's another experiment. I didn't use a Pin Light layer(that's been giving me a lot of trouble lately), but I added in a Screen layer. Screen is supposed to act like the opposite of Multiply. I'll try importing this HF into Wilbur and see what happens. Maybe tomorrow.

I also tried some naming and labelling.

Anyway, here is Al Burphaban Island...

EDIT: I added in another image showing the unedited noise. It would be interesting to overlay them...