Voxel rendering Motion blur bug??

anon52888474 · September 23, 2009, 3:40pm

with max 2009 and Krakatoa 1.5, I try to render the particle with the voxel mode and motion blur, but the artifact still there in my alpha
I would like to find a solution in 3d before add a RS motionblur in comp to fix the problem!
Look at my images,
need help please.

anon58450692 · September 23, 2009, 4:25pm

What are your motion blur settings when rendering this? Can you be more clear as to what the artifact is exactly? I’m not sure what is wrong from looking at this image. Perhaps it would be useful to post a render of the same frame without motion blur.

anon52888474 · September 23, 2009, 5:10pm

for sure:!
krakatoa MB setting image and the same render frame without MB!

the Krakatoa MB on the voxel rendering look like add Horizontal pattern on the MB.

best ragards
Gaetan

anon52888474 · September 23, 2009, 5:53pm

the last images is the same particle but with geo cube and rendering with MentalRay + MB

anon58450692 · September 23, 2009, 6:56pm

What are your voxel size settings? You should compare the results of halving (or quartering) the voxel size and doubling the voxel filter radius. You might see this sort of behavior if your voxels are really big on-screen, and your filter radius is 1.

anon52888474 · September 23, 2009, 7:06pm

Hi!
The radius is 0.6
the filter is 1

anon52888474 · September 23, 2009, 8:14pm

thanks problem solve!
it slower but it’s ok
i put the radius down a bit like 0.4
and I double the filter!
Cool!

Chad · September 24, 2009, 6:13pm

So what causes this? I see it in a lot of voxel renders, but don’t see an obvious cause for it.

anon58450692 · September 25, 2009, 4:01pm

Here’s the frame work for understanding the problem:

There is a 3D lattice of sample points (ie. voxels) in space separated by “Voxel Size” units in each dimension.
A particle’s values are filtered onto these voxels using its “Position” channel, and a filter with radius “Voxel Filter Radius”.
As a caveat to number 2, the weights from the filter are normalized so that the entire energy of the particle is present in the voxel field.

If you have a single particle with “Voxel Filter Radius” of 1, this essentially means that the particle contributes to exactly the 8 nearest lattice points. This is essentially a trilinear interpolation. If this particle is assigned a high density (which most people seem to do when testing this), all 8 lattice points will get a density that is fairly high, no matter where the particle lies in between.Now, as we move this particle across the boundary of where it contributes to a different 8 set of lattice points, you will see an apparent jump of the cube (ie. aliasing).

This is negated by picking a higher filter radius, so the relative size of a voxel versus the size of the particle allows for a better reconstruction of the density function.

Is this relatively clear, or should I try to make some illustrations?

Chad · September 25, 2009, 4:48pm

Not the aliasing, I get that. I mean the darkening of the voxel edges. It looks like shadow map problems, which is what I assumed it was until I started doing unlit emissive renders and got the same problems.

Chad