motion blur

Started by digitalguru, March 03, 2018, 05:55:21 AM

Previous topic - Next topic

digitalguru

Hi all,
I'm just doing some experiments composting cg elements (rendered from Maya) with Terragen renders and noticed some artefacts when rendering a simple terrain with 3d motion blur (motion blurred edge of terrain against sky) This goes away when disabling "defer atmo/" etc) -  what are the pitfalls of turning this off? I've a rough idea this enables the ray-tracer to adaptively sample atmosphere?

p.s - should add the artefacts are entirely with the Terragen render, nothing to do with the composite

Matt

Disabling defer atmo/cloud will enable micropoly rendering of atmosphere and clouds (except where they appear over ray traced objects). This will probably increase noise in the sky. You can usually compensate for this by increasing atmo samples and/or cloud ray march quality, but it is sometimes less efficient to render this way and this difference  increases as improvements are made to the atmo/cloud renderer in recent versions and in future. Micropoly cloud edges will be more ragged and sensitive to Micropoly Detail, but they can still look pretty good in many cases. Parallax motion blur in clouds is only possible with defer atmo/cloud (and 3D motion blur), but "nodal" camera rotation blur works perfectly fine with micropoly rendering.

Lots more detail here:

https://planetside.co.uk/wiki/index.php?title=Ray_Trace_Objects_and_Defer_Atmosphere/Cloud

Matt
Just because milk is white doesn't mean that clouds are made of milk.

digitalguru

#2
thanks for that, I had a hazy idea of what defer atmos does, but I see now

I was making a test for motion blurred terrain close to camera:

defer atmos ON:
[attachimg=1]

defer atmos OFF:
[attachimg=2]

So defer off renders everything with the scanline micropoly renderer and you lose the artefacts which I guess are revealed when defer is on and the hybrid scanline / raytrace crosses over at the leading edge of the motion blur.

So would it be right to say to reduce the artefact would either be to turn defer off and render everything scanline (would need more atmos samples etc) or raytrace everything (very expensive)?

I guess the best solution would be to render separate terrain and atmos passes?


Matt

Quote from: digitalguru on March 05, 2018, 05:31:55 PM
So would it be right to say to reduce the artefact would either be to turn defer off and render everything scanline (would need more atmos samples etc) or raytrace everything (very expensive)?

Yes.

Quote
I guess the best solution would be to render separate terrain and atmos passes?

Compositing separate passes doesn't produce the correct result if each pass has motion blur (because it doesn't account for time-varying occlusion), so no. But it's hard to notice the difference, and sometimes it even looks better than physically correct motion blur. If this gives you shorter render times or more flexibility in comp or rendering, then you should do it IMO.

Is the motion blur problem in your actual shot more serious than in the images you posted?

Matt
Just because milk is white doesn't mean that clouds are made of milk.

digitalguru

QuoteIs the motion blur problem in your actual shot more serious than in the images you posted?

Probably not, I'm planning an animation where the camera is close to the terrain, comping in an element from Maya (Arnold) and the terrain will be occluding the Maya render at speed with a decent amount of motion blur.

I thought motion blur might be an issue, which is why I made this test and then saw these artefacts. In the final animation however, I might not see extreme motion blurred terrain against any atmosphere which is why I'm parking the test right now, but still it was good to get my head around the various aspects of the hybrid renderers.

Quotetime-varying occlusion

What is that? Do you mean by separating out terrain and atmospheres you would be essentially rendering with scanline (terrain) and raytrace (atmosphere) and the motion blur / occlusion will be different?


Matt

#5
Quote from: digitalguru on March 07, 2018, 06:43:25 AM
Quotetime-varying occlusion

What is that? Do you mean by separating out terrain and atmospheres you would be essentially rendering with scanline (terrain) and raytrace (atmosphere) and the motion blur / occlusion will be different?

I don't know if there's a technical term for it, but in photography if you have a foreground object which is blurred and a background object which is also blurred (either by motion blur or by being out of focus) it's possible for the background object to appear sharper wherever it is partly occluded by the foreground object. This is because less of the background blur is visible through the foreground blur, effectively bringing the background object back into focus. The foreground object acts upon the background in the same way that squinting with your eyes improves focus. It looks weird when you see it in a photo, but it really happens. But because it looks weird it might not be something you care to simulate. If you do an all-in-one 3D render you will get this effect, but if you comp elements separately you won't.

I would post a photograph showing this phenomenon but I haven't been able to find one in a brief search.

In the specific case of motion blur, I call it "time-varying occlusion" because the exposure time window onto the background depends on when the foreground object occluded the background, and that varies during the exposure window.

Matt
Just because milk is white doesn't mean that clouds are made of milk.

digitalguru

#6
That's interesting, I haven't come across that before. In VFX it's rare for a shot to rendered all in one, so I'd imaging the potential for that to happen is quite high, though it sounds like a very particular set of circumstances for that effect to be visible. I'll have to ask some compositor friends of mine if they've been asked to simulate it.

QuoteI call it "time-varying occlusion"

I see what you mean, I guess it could be fixed by varying the motion blur position, but for a fast moving shot I'd say it would be hard to spot.


Matt

Quote from: digitalguru on March 08, 2018, 05:47:34 AM
I see what you mean, I guess it could be fixed by varying the motion blur position, but for a fast moving shot I'd say it would be hard to spot.

It would have to be varied per pixel. But yeah, it's hard to spot and I don't think anyone would know or care.

Matt
Just because milk is white doesn't mean that clouds are made of milk.

bobbystahr

Thanks guys, an interesting and informative thread...it's snowing here so I'm catching up on reading....heh again
something borrowed,
something Blue.
Ring out the Old.
Bring in the New
Bobby Stahr, Paracosmologist