Mantra DoF Sampling

[TheDude]

So I've noticed that with enabling depth of field in mantra, the actual camera controls are pretty intuitive, and easy, and do what you expect. What I've noticed though is that the renderer seems to 'blur' the correct area in depth that it should, but as it increases the blur (increased distance from focal distance), it becomes progressively more noisey. This is pretty unfortunate because without DoF, you'll set up your render settings to make sure everything looks nice (the threshold between time and noise), then turn on DoF, setting up camera, and only the area in focus will be at the threshold of acceptable render quality, as everything else gets more noisey.

Is there anything anyone can give any insight to this? I know motion blur has the progressive sampling thing to allow mantra to figure out where to focus more on based on its motion blur, so as to not waste time extra sampling on something static -- does anyone know of a trick like this for depth of field in mantra?

[anim]

motion blur and DOF use the same Motion Factor Property to optimize(reduce) shading quality in highly blurred areas

read the help for more details about motion factor http://www.sidefx.com/docs/houdini12.5/props/mantra#id283036

but to your question, to make your blurred areas less noisy you'd need to increase pixel samples until you like the blurred areas, then you can play with other settings to optimize/compensate for increased pixel samples

[TheDude]

Thanks for the reply. Despite Motion factor becoming an enabled parameter when DoF is switched on, it actually has no effects on render time or the render itself (2 pics attached motion factor 1 and 0). It says that it optimizes blurred areas, but I think that's referring to only blurred due to movement. Thanks for the help though. Perhaps in the future they'll enable a new parameter to help compensate for the extra noise with DoF.

[anim]

I didn't say that that property will help you with your problem, just pointed to the help page where it is explained as you've had incorrect information before, it doesn't change sampling of the image, just can decrease quality of the procedural geometry like displacements or render time subdivisions or NURBS tesselation etc, so if you are not using any of that, then no change will happen. And it works with DOF as well, not only motion blur

so for now only the pixel samples to improve motion blur or DOF

[dennis.albus]

I think what you are talking about would be an adaptive pixel sampler like VRAYs adaptive DMC. I've created an RFE for that just a moment ago as that is something I'm wishing for in mantra for a long time. This would speed up renders with lots of MB and DOF quite a bit.

-dennis

Edited July 18, 2013 by dennis.weil

[symek]

Thanks for the reply. Despite Motion factor becoming an enabled parameter when DoF is switched on, it actually has no effects on render time or the render itself

As anim said, motion factor has an impact on procedural geometry generation. Most noticeable in micropoly modes, when it can significantly reduce the number of shaded samples.

[lukeiamyourfather]

If you tell Mantra to do more samples it will reduce the noise in the out of focus areas. If it would help I can post a scene file for comparison. Unfortunately though Mantra seems to render depth of field very slowly, maybe we'll see that improve in the next Houdini if you submit RFE?

Edited July 18, 2013 by lukeiamyourfather

[Guest tar]

FWIW. Doing some Dof research and this is updated in H13: 

 

Decorrelate Depth of Field from Motion Blur Samples

 

http://www.sidefx.com/docs/houdini13.0/props/mantra