Dinosaur spikes

[robcat2075]

My grandest material essay so far... bony stegosaurus style-plates with contrasting tiger-striped Dino-Hyde© in between.

 

 

 

 

This is one color material...

 

 

and one displacement material (the skin detail is very faint but it's there)...

 

 

and no maps!

 

 

 

a turn-around:

platesAllDispH264.mov

 

Yeah, it's got some pops.

[mtpeak2]

Looks pretty cool.

 

If you lower the displacement some, does the popping go away? What if you made the spikes rounder at the base, so it's less oval?

[jason1025]

Very cool. The quality of your renders are improving. It looks photo real to me.

[robcat2075]

If you lower the displacement some, does the popping go away? What if you made the spikes rounder at the base, so it's less oval?

 

The popping is always there. Lower displacement makes it seem less because the pop travels over a shorter distance. I can't tell that the shape has anything to do with it.

 

I think it has something to do with the angle that the original surface is seen by the camera or possibly the angle that two details on the original surface happen to make to the camera. If the exact circumstances could be identified a fix could probably be had.

 

None-the-less this displacement material stuff has potential, I think.

[PF_Mark]

Looks more like a lighting issue to me?

[NancyGormezano]

Looks more like a lighting issue to me?

 

me too...perhaps try rendering the model with flat shaded option - to test if the "displaced geometry" isn't causing the popping - try with different light types? Maybe shading computations aren't precise enough for displacement?

[robcat2075]

Tell me more about the lighting theory.

 

Here's a test of the bitmapped displacement model lit from two different directions. The popping still seems to be happening on the same frames.

 

LightDirectionComparison.mov

[3DArtZ]

what version did you do this in?

[NancyGormezano]

Tell me more about the lighting theory.

 

Hmmm...yup...your example looks like the computed displacement is popping, rather than any computed shading anomalies.

 

I ran a short test - with a displacement decal (24 bit tga) at 1000%, 4 pass, - no conclusions on my part - but scintillation seems to occur in areas of high contrast (upper left, lower right), as well as in those areas that resolve to small number of pixels. Maybe more passes might help. I have no idea what's the problem. Adding a bump map - didn't alleviate.

disp2withbump4mph264.mov

[robcat2075]

It's not a new problem really. It's been around for as long as we've had this new displacement. But I still think it's about the angle the original patch is seen by the camera.

 

Here's a test with the same rotation seen from two different angles.

 

platesAngleComparison.mov

 

 

 

what version did you do this in?

 

This is v15i

[brainmuffin]

It looks to me like no matter what, it's always the same plates popping. What is it about those plates? Are they reaching pure white on your displacement map? If they are, maybe you should bring the highlights down a little on the map, then boost the percentage of displacement... You could also experiment with the resolution of the displacement map, maybe a higher res would pop less?

[robcat2075]

Are they reaching pure white on your displacement map? If they are, maybe you should bring the highlights down a little on the map, then boost the percentage of displacement...

 

That was an idea worth a try. Here the levels top out a 180 instead of 255...

 

PlatesShallowerTestMP4.mov

 

Still popping though...

 

 

 

 

You could also experiment with the resolution of the displacement map, maybe a higher res would pop less?

 

I have tried higher res and that didn't help. And the test with the material displacement would be infinite res and that still had popping too.

[HomeSlice]

What about if you increase the Octaves in your displacement material?

... just brainstorming...

[robcat2075]

What about if you increase the Octaves in your displacement material?

... just brainstorming...

 

Octaves increases the level of detail in the greyscale pattern. But it doesn't seem any less pop vulnerable, based on my previous test using displacement material for the landscape flyover.

 

However, the example above on the white model is bitmapped displacement not a material like the example at the very top.

[HomeSlice]

However, the example above on the white model is bitmapped displacement not a material like the example at the very top.

What about using a 16-bit grayscale image for displacement instead of 8-bit?

I know mesh resolution should not matter, but have you tried increasing it?

I think it has something to do with the angle that the original surface is seen by the camera or possibly the angle that two details on the original surface happen to make to the camera.

You're probably right, but I know nothing of how the renderer works.

[robcat2075]

However, the example above on the white model is bitmapped displacement not a material like the example at the very top.

What about using a 16-bit grayscale image for displacement instead of 8-bit?

 

It's already an exr, which is to infinity and beyond 16 bit . I tried 8 bit maps first but they have too much banding when you get out to 1000% displacement levels.

 

 

I know mesh resolution should not matter, but have you tried increasing it?

 

I haven't tried on this one because on my previous experiments it seemed not to change anything.

 

I may just have to wait for Martin to get excited about this feature again.

[NancyGormezano]

I found that there are less anomalies when the image used for displacement has smoother transitioning.

 

Used same image (800 x 800 24 bit tga) progressively blurred (with gaussian 20 pixel radius) for displacement (and color - cylindrical map 6 x 1).

 

The right most vase has the most problems obviously - too sharp transitions.

4vasetestbiggercolor50AIh264loop.mov

[robcat2075]

But smooth spikes are no fun!

[robcat2075]

Here's the material displacement model again , with a smoother hump pattern.

 

It's still popping, but as I watch it closely it's not like the humps are getting higher or lower, it's more like a patch on the top is disappearing and reappearing.

 

platesAllNonEuclidH264.mov

[itsjustme]

It seems like when one of the spikes is mostly obscured by another spike is when the popping happens...I wonder if the calculations clash somehow when that happens. This is a great test, Robert!

[NancyGormezano]

It's still popping, but as I watch it closely it's not like the humps are getting higher or lower, it's more like a patch on the top is disappearing and reappearing.

 

yup - I wonder if you smooth your pattern more if that would help, but I suspect there will always be anomalies for some view angle.

 

Too bad can't bake the "displaced" geometry in the model - rather than have it computed on the fly (or that's what looks like is happening)

[jason1025]

Thats disappointing.

[Rodney]

Does it do the same thing if the camera circles around the dino instead of the dino rotating?

[3DArtZ]

that's actually a pretty good question you have there rodney.

I'd like to see what happens with the dino staying still and the camera moving around it.

Mike Fitz

[robcat2075]

Does it do the same thing if the camera circles around the dino instead of the dino rotating?

 

hmmm... I'll give that a try when I reconstruct my stuff for this. These little experiments are among the things lost with my hard drive.