View previous topic :: View next topic |
Author |
Message |
Jeroen Site Admin
Joined: 07 Aug 2004 Posts: 5332 Location: The Netherlands
|
Posted: Thu May 17, 2007 5:21 pm Post subject: Subdivision function |
|
|
People want subdivision in DeleD so... you ask, we deliver. I'm working on subdivision right now. Early screens:
Original object (created using the new Lathe function) on the left: 324 polygons
After one subdivision: 1272 polygons
After a second subdivision: 5088 polygons
I really like playing with the combination of Lathe and Subdivision!
Original object on the left: 22 polygons
After one subdivision: 88 polygons
After a second subdivision: 352 polygons
Still got some work to do but subdivision will be available in the next DeleD PRO version (1.7). |
|
Back to top |
|
|
Daaark DeleD PRO user
Joined: 01 Sep 2004 Posts: 2696 Location: Ottawa, Canada
|
Posted: Thu May 17, 2007 5:44 pm Post subject: |
|
|
Very nice.
The jumps seem to be huge. You can't just double the polycount, instead of 4x? (All modellers do this, nothing against your implementation) _________________
|
|
Back to top |
|
|
Jeroen Site Admin
Joined: 07 Aug 2004 Posts: 5332 Location: The Netherlands
|
Posted: Thu May 17, 2007 5:46 pm Post subject: |
|
|
Daaark wrote: |
Very nice.
The jumps seem to be huge. You can't just double the polycount, instead of 4x? (All modellers do this, nothing against your implementation) |
That depends on what subdivision scheme is being used. I've used Catmull-Clark subdivision which roughly generates 4 polygons for each vertex in the original object. Most other subdivision scheme's produce a more or less simular result. Doubling the polygon... I wonder which subdivision scheme comes close to that... Do you know?
Edit: I actually think more refinement is possible by merging polygons after applying the subdivision. I might just investigate that too... perhaps... |
|
Back to top |
|
|
Mr.Fletcher DeleD PRO user
Joined: 07 Aug 2004 Posts: 1772 Location: Germany
|
Posted: Thu May 17, 2007 5:55 pm Post subject: |
|
|
That looks brilliant! _________________ Behold! The DeleD Wiki! Please help us expanding it
DeleD on IRC |
|
Back to top |
|
|
Daaark DeleD PRO user
Joined: 01 Sep 2004 Posts: 2696 Location: Ottawa, Canada
|
Posted: Thu May 17, 2007 6:13 pm Post subject: |
|
|
Jeroen wrote: |
Edit: I actually think more refinement is possible by merging polygons after applying the subdivision. I might just investigate that too... perhaps... |
No need, I was just curious.
Does this only work on quads? _________________
|
|
Back to top |
|
|
Jeroen Site Admin
Joined: 07 Aug 2004 Posts: 5332 Location: The Netherlands
|
Posted: Thu May 17, 2007 7:04 pm Post subject: |
|
|
Daaark wrote: |
Does this only work on quads? |
Nope, it works on arbitrary meshes so triangulated meshes are no problem either. The algorithm does produce quads only though. |
|
Back to top |
|
|
jwatte DeleD PRO user
Joined: 26 Apr 2006 Posts: 513
|
Posted: Fri May 18, 2007 12:06 am Post subject: |
|
|
Now, if there was a good "optimize" button to un-subdivide too :-) |
|
Back to top |
|
|
Jeroen Site Admin
Joined: 07 Aug 2004 Posts: 5332 Location: The Netherlands
|
Posted: Fri May 18, 2007 5:53 am Post subject: |
|
|
jwatte wrote: |
Now, if there was a good "optimize" button to un-subdivide too |
Yes, I've been thinking about that as I've seen that in other modellers too. I wonder... could an arbitrary object that hasn't been subdivided yet be un-subdivided? Interesting stuff... |
|
Back to top |
|
|
Vijchti Member
Joined: 16 Aug 2006 Posts: 250
|
Posted: Fri May 18, 2007 7:54 am Post subject: |
|
|
...It's like you know exactly how to turn me on. _________________
"Psst, Here's a secret...Your last mortal thought will be,
'Why did I take so many days - just like today - for granted?'" |
|
Back to top |
|
|
Jeroen Site Admin
Joined: 07 Aug 2004 Posts: 5332 Location: The Netherlands
|
Posted: Sun May 20, 2007 1:04 pm Post subject: |
|
|
I finished up the subdividing routines (for now anyway ) and implemented another subdividing scheme: Doo-Sabin. Take a look:
Left object: original
Middle object: Catmull-Clark subdivision applied one time
Right object: Doo-Sabin subdivision applied one time
Left object: original
Middle object: Catmull-Clark subdivision applied two times
Right object: Doo-Sabin subdivision applied two times
As you can see, the Catmull-Clark scheme produces a slightly rounder object while Doo-Sabin follows the curves of the original object a little more. Both ways have their uses, of course. Amount of polygons in both subdivided objects is almost the same.
The smoothing algorithms are found in the Lowpolygon Command frame (although you can hardly call this lowpolygon anymore ) and can be used in the PRO version. |
|
Back to top |
|
|
granada Team member
Joined: 07 Aug 2004 Posts: 1955 Location: England
|
Posted: Sun May 20, 2007 5:29 pm Post subject: |
|
|
Cool, nice work there jeroen .
Dave _________________ AMD Phenom(tm)IIx6 1090t Processor 3.20 GHS
8.00 GB memory
Windows 7 64 bit
Nvida Geforce GTX 580 |
|
Back to top |
|
|
Daaark DeleD PRO user
Joined: 01 Sep 2004 Posts: 2696 Location: Ottawa, Canada
|
Posted: Wed May 23, 2007 4:01 pm Post subject: |
|
|
Are texture cordinates maintained? _________________
|
|
Back to top |
|
|
espritcool Member
Joined: 02 Aug 2006 Posts: 91
|
|
Back to top |
|
|
Jeroen Site Admin
Joined: 07 Aug 2004 Posts: 5332 Location: The Netherlands
|
Posted: Wed May 23, 2007 6:43 pm Post subject: |
|
|
Texture coordinates are recalculated because additional vertices are inserted.
If your object can still be cut into convex pieces, I see no reason why the Torque exporter (and Torque itself) wouldn't like those objects. |
|
Back to top |
|
|
jwatte DeleD PRO user
Joined: 26 Apr 2006 Posts: 513
|
Posted: Wed May 23, 2007 9:47 pm Post subject: |
|
|
Quote: |
Texture coordinates are recalculated |
The right thing to do is probably to generate texture coordinates for the new vertices using barycentric interpolation of the old vertices on the edge (or face) being split. |
|
Back to top |
|
|
|