• Welcome to World Machine Community. Please login or sign up.
May 21, 2019, 03:51:16 am


Read the Development Diary for an inside look at World Machine's progress!

Terrain Synthesis from DEMs

Started by Alexander, March 18, 2007, 02:03:38 pm

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.


March 18, 2007, 02:03:38 pm Last Edit: March 18, 2007, 04:15:41 pm by Alexander
This paper is an interesting read and I couldn't help but imagine how this technique could be applied to the already rich World Machine.



"Every time you read or write beyond the end of an array, somewhere a fairy dies."
Richard Buckland


March 19, 2007, 02:53:06 pm #2 Last Edit: March 19, 2007, 02:55:29 pm by monkschain
Hey what a great idea- transcribe mountains as curvilinear networks from a topo map of Middle Earth - now why didn't we think of that!  :lol:

The terrain sampling goes one step further of course- something I spoke to Joe Slayton about last year actually- well it was Joe who brought it up- that's amazing- have you seen that terrain!!  YE GADS :shock:
Triple cool rating  8)  8)  8)



Yea, the terrain sampling is what I considered to be the meat of this paper and is what interested me the most. That combined with the features that World Machine offers currently would make for some mind-blowing software.


Absolutely- fantastic find btw Alex- that paper blew my mind. :D
I was thinking if you hooked up the new tools for user input that are out there and coming on line: Wm's LM for eg- but then use a real world terrain to sample from... phew! I do think that the real world terrain sampling is another major piece of the jigsaw for the 'dream editor' :)



Yea, I need to take a closer look at the paper and try to determine how explicitly they explain their process. It Stephen doesn't want to add this feature as part of the application, we'd still probably be interested in implementing it (as a plugin?).

Not sure how it would best fit into the currently WM workflow, as an input or a device which applies the particular look of a DEM file you've loaded to an already existing terrain anywhere in the pipeline. Thoughts?


Hi guys,
    First, thank you all very much for the emails and comments about this paper. My name is Howard Zhou and I am one of the authors. I am a long time admirer of the work that has been done here, and I am really glad that our work can help spur some interesting discussions in the forum. I especially like the terrain sampling idea because I also think it's a much better way to utilize the gazillion bytes of DEMS we have. Maybe in the future, we can go so far as to try reverse engineering on a real terrain heightmap to create a procedure model (erosion model for instance).
    If you guys are interested in the actual implementation details or limitations of the system, I will be happy to discuss them here. Feel free to ask any question and I will try my best to answer it.



March 23, 2007, 10:08:15 am #7 Last Edit: March 23, 2007, 01:12:22 pm by monkschain
Hey all  :), Stephen has really managed to do some amazing things with Devices- I mean who would have thought that you could implement something like WM Pro's Layout mode with a device? I see your point though Alex, LM does operate as a completely new 'mode'.
Imo, hooking this functionality up to the new tools coming along in WM Pro will really be the icing on the cake. All power to your elbow.  I think you might use the shapes in Layout mode. They could serve as your input tool (mask)- much like Howard's user input. The synthesis device would surely be a Combiner of sorts?- combining input from two terrains- the real dem and the layout generator perhaps?
(There's also some (more general) discussion going over here guys if you're looking for ideas.) :)



Quote from: howardzzh on March 22, 2007, 08:33:25 pm
... My name is Howard Zhou and I am one of the authors. ...

Excellent paper!
I am (a WM user and) the developer of a 3D Heightmap Editor application specifically for video game studio licensing only (not for public use), and there are a number of terrain synthesis elements that I am integrating into the new version of the software currently under development, including one that is somewhat similar to what you cover in your paper (albeit not quite as in-depth as what you are accomplishing and on a smaller scale as most current video game heightmap requirements are usually much smaller than visualization systems, averaging 128x128 through 1024x1024 samples in total).
I have been using DEMs for many years now in video game heightmap design, and I am working on integrating methods for mixing numerous fractal/noise/spectral synthesis, erosion simulation, real DEM data, and DEM-based geological feature libraries, with a full 3D spline and mask based interface.
The DEM feature library system I am working on includes a (currently) small number of common real-world DEM geological feature data patches that are then rotated and blended to follow the 3D splines for mountains, rivers, etc. (not all of the bugs are worked out yet).
This is the first time I have seen your paper, and I find your techniques very interesting.
DGUnreal Lead Designer and Engineer
Game Designer/Programmer since 1982


Out of curiosity, is this a real-time solution that you're offering?


Quote from: Alexander on March 29, 2007, 09:13:18 am
Out of curiosity, is this a real-time solution that you're offering?

I assume that this was directed at me?   :)
I didn't mean to hijack the WM forums...

As real-time as I can get it, depending on the specific create/edit/modify function chosen, the working size of the heightmap, and the power of the computer system...
I am not to that point in the application programming yet, but I am looking at multi-threading to allow for real-time heightmap visualization and deformation during object/spline movement etc.
The interface and methodology are quite a bit different than WM, Leveller, Terragen, etc. (think 3DS Max plus UnrealEd).  So it has a different target market to a large degree than the other applications I just mentioned.
The software will be (as close to) retail-quality (as possible), as I am a long-time professional software developer, so the full development cycle will be some time yet.
I have some of the largest game companies and top level designers already on my beta tester team.
The Basic version is available to Unreal Engine Licensees only, the Full version is available only to game studios, no public release will be available.

Any other questions regarding this should probably be taken off of these boards (PM or Email).  ;)
DGUnreal Lead Designer and Engineer
Game Designer/Programmer since 1982


Sounds interesting.  :D
I would like to hear Stephen's take on this technology for WM still!


I'd also be really interested to hear Stephen's take on this. I think getting this functionality into WM would be a real step change- especially when hooked up with the new vector tools in the upcoming WM Pro. Howard seemed to be interested in vectors as a mode of input for the user-sketched feature map. That suggests to me that hooking the two up is doable.
I can see a potential workflow between real world GIS vector data and Terrain Synthesis via vectors. If you're interested, some ideas here.
Alex, have you written other WM plugins? Do you code as part of a team?




Neither, I am just a consumer beta-testing the product for use with a game project I am involved with. But I do have access to development resources if need be. Obviously it would be ideal if such a feature were implemented as part of WM rather than a plugin though.


In terms of my take on the terrain synthesis from DEMs:

It's a very impressive paper with quite good looking results. Back a few years ago when the first papers on Image Analogies came out, I started wondering what would happen if you applied the same idea to heightfields -- and this is the answer.

With that said, without playing around with the tech it's hard for me to say more. In my mind the principle weakness of the method is the same as its strength -- it's reliance on DEM data for features. If you have geospecific areas you're trying to model (ie " I want the Rocky Mountains, but in my own world") it seems to work quite well.
-- Stephen Schmitt
- Creator of World Machine


"If you have geospecific areas you're trying to model (ie " I want the Rocky Mountains, but in my own world") it seems to work quite well."

That is EXACTLY the type of thing that would make life easier for my studio.

"I want to be able to design a map that resembles a certain mountainous region in France without tweaking for ages or doing an ungodly amount of reference analysis."


I think that type of thing may actually lie at the root of *many* people's desires for semi-explicit terrain, which IMO is really WM's whole paradigm (as opposed to fully explicit terrain like Leveller, etc. that allow painting-type interaction). Although WM may have satisfied many people's desires to create realistic artificial terrain, I can remember since the beginning essentially that many people have come around asking how to reproduce this or that type of real-world terrain. If sampling can be used to create suitably unique, semi-random and still realistic variations of real-world data with infinite variations then I think there would be a lot of takers.

- Oshyan

Quick Reply

With Quick-Reply you can write a post when viewing a topic without loading a new page. You can still use bulletin board code and smileys as you would in a normal post.

Warning: this topic has not been posted in for at least 120 days.
Unless you're sure you want to reply, please consider starting a new topic.

Note: this post will not display until it's been approved by a moderator.

Please leave this box empty:
Type the letters shown in the picture
Listen to the letters / Request another image

Type the letters shown in the picture:

Shortcuts: ALT+S save/post or ALT+P preview