Subdivision + displacement test

BbB · Post by **BbB** » Wed May 14, 2008 8:07 am

Really big displacement with 8bit images always shows this stair effect, regardless of resolution. How much it shows to the naked eye depends on how "displaced" the surface is. I guess displacement is best used sparingly and any huge displacement is better modelled in any case, for render speed and efficiency purposes.

drBouvierLeduc · Post by **drBouvierLeduc** » Thu May 15, 2008 2:48 am

Mmmmh, checking my hungarian/english dirty phrasebook... Got it :
"My nipples are exploding with delight".

zsouthboy · Post by **zsouthboy** » Thu May 15, 2008 3:45 am

drBouvierLeduc wrote:Mmmmh, checking my hungarian/english dirty phrasebook... Got it :
"My nipples are exploding with delight".

I love the reference.

BbB · Post by **BbB** » Thu May 15, 2008 4:28 am

Displaced nipples... That's an idea... Mmmh.

(Docteur!!!! Where have you been? What about that villa you never finished. It still haunts my dreams.)

Post by **OnoSendai** » Fri May 16, 2008 2:15 am

Testing an adaptive subdivision heuristic based on curvature:

The material is a hack to the diffuse texture that gives each triangle a different colour

PureSpider · Post by **PureSpider** » Fri May 16, 2008 2:21 am

w00t this looks good (and funny

)
Give us the hacked mat! XD

dougal2 · Post by **dougal2** » Fri May 16, 2008 2:35 am

amazing. do we get info output on the number of faces in the mesh before and after subdivision ?

Post by **fused** » Fri May 16, 2008 2:45 am

great! wish c4d had such displacement...

Post by **OnoSendai** » Fri May 16, 2008 2:45 am

probably, Dougal2.

BbB · Post by **BbB** » Fri May 16, 2008 2:52 am

Looking sweet.

From the user's point of view, will the choice between modelling the displacement or letting Indigo handle it be a trade-off between memory consumption and rendering efficiency/speed?

Pibuz · Post by **Pibuz** » Fri May 16, 2008 3:29 am

From the point of view of sketch users, i would definetely go for letting indigo calculate, for there is no displacement modeling tool in Sketchup!

drBouvierLeduc · Post by **drBouvierLeduc** » Fri May 16, 2008 3:44 am

Oooh, sweet. Reminds me of that thread at BA's forum (micropoly script & tests).
Now I don't know if this has anything to do with your method, but it sure looks the same.
There's an option to subdivide only what's visible to camera btw (to save memory I presume), but I don't know if that can apply to an unbiased renderer.

Ok I swear I'll never attempt to have some technical discussion ever again !
back to my 3d puppets and stuff.

Post by **OnoSendai** » Fri May 16, 2008 3:58 am

Yeah, it's somewhat similar to that, dr.

Visibility computations can be tricky, for example, what if you are looking through a pane of glass?

BbB: not sure yet

dougal2 · Post by **dougal2** » Fri May 16, 2008 4:07 am

or looking into mirrors and other reflective surface ?

Post by **OnoSendai** » Fri May 16, 2008 4:13 am

indeed, Dougal.

Right now I'm pondering whether it would be possible to turn Indigo into some kind of REYES style raytracer, e.g. to abandon shading normals altogether and rely on subdivided triangles to generate sufficiently smooth geometry.

This would be nice because, as you may know, shading normals are a disgusting un-physical hack.

The main problems with this idea are mem usage and rendering performance.

However, quite a lot depends on what kind of subdivision resolution would be required. Anyone have any thoughts on that?

Attached are some tests with only geometric normals (no shading normals)

Subdivision + displacement test

Who is online