[Algorithms] 2D Side Terrain generation

Discussion:

Joel B

2011-08-09 02:22:03 UTC

Hello, I'm looking for algorithms for generating 2d terrain from a side view, like the old Defender game from the 80's, except with mountains grass, rocks, lakes etc instead of just lines.

Here is an example of what I have done in Photoshop:

Loading Image...

Anyone on this list have some suggestions? It doesn't have to be real time, although that would be interesting too.

- Joel

Brennan Rusnell

2011-08-09 03:39:48 UTC

Permalink

I'd recommend using Ridged Multifractals:

https://www.internal.pandromeda.com/engineering/musgrave/unsecure/S01_Course_Notes.html
http://www.ypoart.com/downloads/Musgrave.htm
http://vterrain.org/Elevation/Artificial/

This is a great book on the topic:
http://books.google.com/books/about/Texturing_modeling.html?id=bDlSJd8GfMcC

Hope this helps.
--
Brennan Rusnell

Post by Joel B
Hello, I'm looking for algorithms for generating 2d terrain from a side
view, like the old Defender game from the 80's, except with mountains grass,
rocks, lakes etc instead of just lines.
http://img855.imageshack.us/img855/8125/examplelp.png
Anyone on this list have some suggestions? It doesn't have to be real time,
although that would be interesting too.
- Joel
------------------------------------------------------------------------------
uberSVN's rich system and user administration capabilities and model
configuration take the hassle out of deploying and managing Subversion and
the tools developers use with it. Learn more about uberSVN and get a free
download at: http://p.sf.net/sfu/wandisco-dev2dev
_______________________________________________
GDAlgorithms-list mailing list
https://lists.sourceforge.net/lists/listinfo/gdalgorithms-list
http://sourceforge.net/mailarchive/forum.php?forum_name=gdalgorithms-list

Joel B

2011-08-09 04:46:02 UTC

Permalink

Thanks very much for the suggestions!

- Joel

Marc Hernandez

2011-08-09 05:34:07 UTC

Permalink

The big revelation for me was that I realized I could use fractals
as input functions to other fractals, as well as composing them from
adding or multiplying.

I tend to start with the standard perlin noise to define the major
continents, then use things like rigid multifractal to build mountains
into it.

Megan Fox

2011-08-09 13:49:44 UTC

Permalink

I'd also recommend Perlin - once you start mixing high or low passed
low-frequency Perlin noise with high/low-passed high-frequency noise, you
get some very, very interesting results.

The key, and what will define your continents and the feel of the resultant
shape, is the number of layers of noise, and how you scale or cut (high/low
pass) each layer.

Hugo Elias's page is an oldy, but it has some really great info on this
topic: http://freespace.virgin.net/hugo.elias/models/m_clouds.htm - that
page is the one that turned me on to how cool noise could be in the first
place (my version of his cloud algorithm:

I realize he's working in a
slightly different space, but those basic techniques can be applied from a
side view just as they can from a top-down view.

Andy Farnell

2011-08-09 14:51:18 UTC

Permalink

An alternative take to random/quasi-stochastic methods I'll mention
an exercise we did with my London undergraduates for a Lunar Lander
terrain. The problem was to make the terrain controllable, in order to
insert plausible landing sites, (or IIRC "defender" well, you might
want to create volatile terrain to nicely hide ground based missle silos)

Rather than starting with a Perlin source and filtering as Megan suggests,
we found a great way was to Fourier construct with about 5 or 6 sine components.
By setting the frequency and amplitude coefficients the terrain could have
a useful "roughness" parameter, with local consistency, without unexpected
dicontinuities (effectively band limiting the terrain within the frame
resolution) and remain well bounded in height. Compared to a pink or
Perlin source and IIR filter it's swings and roundabouts in terms of cost.

- Andy

On Tue, 9 Aug 2011 07:49:44 -0600

Post by Megan Fox
I'd also recommend Perlin - once you start mixing high or low passed
low-frequency Perlin noise with high/low-passed high-frequency noise, you
get some very, very interesting results.
The key, and what will define your continents and the feel of the resultant
shape, is the number of layers of noise, and how you scale or cut (high/low
pass) each layer.
Hugo Elias's page is an oldy, but it has some really great info on this
topic: http://freespace.virgin.net/hugo.elias/models/m_clouds.htm - that
page is the one that turned me on to how cool noise could be in the first
http://youtu.be/kHvJqPxXUXM I realize he's working in a
slightly different space, but those basic techniques can be applied from a
side view just as they can from a top-down view.

Joel B

2011-08-09 20:20:35 UTC

Permalink

Hi Andy, do you have any examples of the algorithm? Is it just a standard Fourier transform and use the waveform output as landscape?

- Joel

Post by Andy Farnell
An alternative take to random/quasi-stochastic methods I'll mention
an exercise we did with my London undergraduates for a Lunar Lander
terrain. The problem was to make the terrain controllable, in order to
insert plausible landing sites, (or IIRC "defender" well, you might
want to create volatile terrain to nicely hide ground based missle silos)
Rather than starting with a Perlin source and filtering as Megan suggests,
we found a great way was to Fourier construct with about 5 or 6 sine components.
By setting the frequency and amplitude coefficients the terrain could have
a useful "roughness" parameter, with local consistency, without unexpected
dicontinuities (effectively band limiting the terrain within the frame
resolution) and remain well bounded in height. Compared to a pink or
Perlin source and IIR filter it's swings and roundabouts in terms of cost.
- Andy
On Tue, 9 Aug 2011 07:49:44 -0600

Post by Marc Hernandez
The big revelation for me was that I realized I could use fractals
as input functions to other fractals, as well as composing them from
adding or multiplying.
I tend to start with the standard perlin noise to define the major
continents, then use things like rigid multifractal to build mountains
into it.
From there I use another perlin noise to add smaller hills and
valleys, and use yet another to define where to place all those.
Here's a 2d example of the output.
http://redmine.alephnought.com/projects/cubit/wiki/World#Pretty-pictures
I didnt have a terrain renderer yet, so I dont know what the
landscape looked like, but I liked the start of it.
For something like you're mentioning, I think I might treat it as a
strip of a standard 2d heightmap. That way you might get some neat
parallax as the ship flys through it.
On Mon, Aug 8, 2011 at 8:39 PM, Brennan Rusnell
https://www.internal.pandromeda.com/engineering/musgrave/unsecure/S01_Course_Notes.html

Post by Brennan Rusnell
http://www.ypoart.com/downloads/Musgrave.htm
http://vterrain.org/Elevation/Artificial/

http://books.google.com/books/about/Texturing_modeling.html?id=bDlSJd8GfMcC

Post by Brennan Rusnell
Hope this helps.
--
Brennan Rusnell

Post by Joel B
Hello, I'm looking for algorithms for generating 2d terrain from a side
view, like the old Defender game from the 80's, except with mountains

grass,

Post by Brennan Rusnell

Post by Joel B
rocks, lakes etc instead of just lines.
http://img855.imageshack.us/img855/8125/examplelp.png
Anyone on this list have some suggestions? It doesn't have to be real
time, although that would be interesting too.
- Joel

------------------------------------------------------------------------------

Post by Brennan Rusnell

Post by Joel B
uberSVN's rich system and user administration capabilities and model
configuration take the hassle out of deploying and managing Subversion

and

Post by Brennan Rusnell

Post by Joel B
the tools developers use with it. Learn more about uberSVN and get a

free

Post by Brennan Rusnell

Post by Joel B
download at: http://p.sf.net/sfu/wandisco-dev2dev
_______________________________________________
GDAlgorithms-list mailing list
https://lists.sourceforge.net/lists/listinfo/gdalgorithms-list

http://sourceforge.net/mailarchive/forum.php?forum_name=gdalgorithms-list
------------------------------------------------------------------------------

Post by Brennan Rusnell
uberSVN's rich system and user administration capabilities and model
configuration take the hassle out of deploying and managing Subversion

and

Post by Brennan Rusnell
the tools developers use with it. Learn more about uberSVN and get a free
download at: http://p.sf.net/sfu/wandisco-dev2dev
_______________________________________________
GDAlgorithms-list mailing list
https://lists.sourceforge.net/lists/listinfo/gdalgorithms-list

http://sourceforge.net/mailarchive/forum.php?forum_name=gdalgorithms-list
--
/// //
------------------------------------------------------------------------------
uberSVN's rich system and user administration capabilities and model
configuration take the hassle out of deploying and managing Subversion and
the tools developers use with it. Learn more about uberSVN and get a free
download at: http://p.sf.net/sfu/wandisco-dev2dev
_______________________________________________
GDAlgorithms-list mailing list
https://lists.sourceforge.net/lists/listinfo/gdalgorithms-list
http://sourceforge.net/mailarchive/forum.php?forum_name=gdalgorithms-list

--
Megan Fox
http://www.glassbottomgames.com/

--
------------------------------------------------------------------------------
uberSVN's rich system and user administration capabilities and model
configuration take the hassle out of deploying and managing Subversion and
the tools developers use with it. Learn more about uberSVN and get a free
download at: http://p.sf.net/sfu/wandisco-dev2dev
_______________________________________________
GDAlgorithms-list mailing list
https://lists.sourceforge.net/lists/listinfo/gdalgorithms-list
http://sourceforge.net/mailarchive/forum.php?forum_name=gdalgorithms-list

Jon Watte

2011-08-14 20:12:17 UTC

Permalink

Andy Farnell

2011-08-14 21:05:17 UTC

Permalink

Oops that was possibly my bad etiquette to fork off there,
I replied to Joel off-list because I wasn't sure about
policy on posting code. As it turns out it was less than
20 lines. Maybe Joel will post back some results later.

Actually it is simple, the synthesis components,
being interpolated lookups of a cosine table (or
math function), are summed and used directly, and if
there's several of them (between 5 and 10 maybe) the
terrain looks okay. Joel sent me a video clip to show
it working.

The next idea I had was maybe to provide some FM feedback
to get a range of quasi-noisy textures

see

Also, do you want heightfield-style terrain (each X has exactly 1 Y) or
overhangs/crags? That may change how you implement the subdivision.

That's an interesting thought.

Andy

On Sun, 14 Aug 2011 13:12:17 -0700
Jon Watte <***@gmail.com> wrote:

Joel B

2011-08-15 04:17:01 UTC

Permalink

Actually I got inspired by Andy's code to start writing code to read and write WAV files, it will be interesting to actually hear the output of his algorithm!

Sent from my iPhone

Post by Andy Farnell
Oops that was possibly my bad etiquette to fork off there,
I replied to Joel off-list because I wasn't sure about
policy on posting code. As it turns out it was less than
20 lines. Maybe Joel will post back some results later.
Actually it is simple, the synthesis components,
being interpolated lookups of a cosine table (or
math function), are summed and used directly, and if
there's several of them (between 5 and 10 maybe) the
terrain looks okay. Joel sent me a video clip to show
it working.
The next idea I had was maybe to provide some FM feedback
to get a range of quasi-noisy textures
see http://youtu.be/Hu9FR_1l0Nc

Also, do you want heightfield-style terrain (each X has exactly 1 Y) or
overhangs/crags? That may change how you implement the subdivision.

That's an interesting thought.
Andy
On Sun, 14 Aug 2011 13:12:17 -0700

From his description, it sounds as if he uses Fourier synthesis to generate

control values for some fractal algorithm. For example, subdivision
algorithms often have "roughness" parameters, which determine how much to
subdivide a segment once it's selected for subdivision; I would imagine
you'd get an interesting-looking terrain by varying roughness in some
locally-smooth but globally-varying way.
Also, do you want heightfield-style terrain (each X has exactly 1 Y) or
overhangs/crags? That may change how you implement the subdivision.
Sincerely,
jw
--
Americans might object: there is no way we would sacrifice our living
standards for the benefit of people in the rest of the world. Nevertheless,
whether we get there willingly or not, we shall soon have lower consumption
rates, because our present rates are unsustainable.

Hi Andy, do you have any examples of the algorithm? Is it just a standard
Fourier transform and use the waveform output as landscape?
- Joel

suggests,

Post by Andy Farnell
we found a great way was to Fourier construct with about 5 or 6 sine

components.

Post by Andy Farnell
By setting the frequency and amplitude coefficients the terrain could

have

Post by Andy Farnell
a useful "roughness" parameter, with local consistency, without

unexpected

Post by Andy Farnell
dicontinuities (effectively band limiting the terrain within the frame
resolution) and remain well bounded in height. Compared to a pink or
Perlin source and IIR filter it's swings and roundabouts in terms of

cost.

Post by Andy Farnell
- Andy
On Tue, 9 Aug 2011 07:49:44 -0600

Post by Megan Fox
I'd also recommend Perlin - once you start mixing high or low passed
low-frequency Perlin noise with high/low-passed high-frequency noise,

you