Render Elements - Coverage ?

[artur.leao]

Hi there. I don't quite understand why Vray supports Coverage (and other channels) in G-buffer, when writing to .rla or .rpf and doesn't support them on Render Elements. Is this a limitation? If not, it would be nice to have those channels in the Render Elements to render them apart. Thanks in advance.

[artur.leao]

No one? This would be a nice improvement for the final 1.5 version of V-Ray...

[vlado]

You have to make a difference between the G-Buffer and the render elements. The G-Buffer is multi-layered: for example, it can store more than one z-depth value per pixel (one value for each object that covers that pixel). The render elements, on the other hand, are just regular images (you have only a single value per pixel). The coverage channel in particular is only meaningful when you use a multilayered G-Buffer. It tells you the contribution to the pixel of each object that covers that pixel. This information cannot be stored in a regular render element, since it only allows one value per pixel.

The scanline renderer also cannot generate the coverage channel as a render element.

Best regards,
Vlado

[Da_elf]

all of this would make alot more sence if i knew how the coverage channel is used. but somehow this has been my most unanswered question on the forum. i bet if i asked what kinda creature richard gere had up his butt that question would be answred and the one about coverage would be forgotten. ive already googles for the coverage channel but came up with nothing

[joconnell]

From what vlado is saying it seems that coverage is used mainly at anti aliased edges and where you have two objects contributing to an anti-aliased pixel, the coverage channel stores information of what percentage each object is contributing to the pixel. It's kind of like a grayscale image where the brightness of the pixel corresponds to how much a particular object has contributed to the pixels in the colour channels so if you have a coverage channel where one layer has a 50% grey pixel, it means that whatever object that particular layer in the coverage channel corresponds to has contributed 50% of the end colour value in that pixel - it's kind of like a matte for each object in a way before the renderer layers all of the objects over each other and spits out a final flattened image.

I'd imagine that it could make some difference to doing stuff like depth of field or motion blur but since it's only adding in maybe a 1 pixel edge at best, it's not going to make a radical enough difference. Whats a lot more useful is something like the render occluded objects functionality although that could lead to huge file sizes in your frames.

[Da_elf]

would it also be good if a render was done with no AA filter then use the coverage to AA in post?

[artur.leao]

That's exactly my point, when doing DOF or MB it always seems to have problems with AA or something like that. When I have an image composed by Elements, at some point there seems to be missing some channels that are present on RPF. But as vlado stated, it looks like the only way of doing this is rendering to RPF... I would still prefer to have all channels in the Elements Tab if possible Cheers

[joconnell]

would it also be good if a render was done with no AA filter then use the coverage to AA in post?

I'd imagine you'd have to render to a format that keeps all of the object data for parts of objects that were covered over by others which would probably be a rather big file - you may end up with a layer for every object in the scene. It'd be a fairly intensive thing to do in a 2d package too I'd imagine and I don't think there's anything that supports it in the first place. Plus if you rendered with no aa there'd be no coverage channel in the first place - if you turn off aa then whatever object is in the front of the render is supplying 100% of the colour value for each pixel which removes the info necessary to make a coverage channel in the first place.

What I reckon you're after from a dof point of view is a type of image that retains the data for objects behind other objects in the scene so that as you blur an image, it's taking the correct colour values from objects that are now spilling into the image via the bleeding that happens with the blur.

I can see why vlado is saying that coverage isn't possible, you'd need to render to something like EXR with each coverage layer being rendered to a seperate EXR channel - it also seems that coverage is something that's only generated by the finished render itself as part of the process of z-depth composing each object into the render.

[Da_elf]

can someone show an example of something like motion blur or DOF before and after the use of coverage channe?

[joconnell]

Won't be me unfortunately - I don't use combustion and I'm not sure how good the after effects support of coverage et al is.

[artur.leao]

There you go, one render is using render elements, therefore in combustion you can only use Box Blur Method when doing DOF, the other one was done by rendering into an RPF and then Gaussian Blur applied, same values where used in both DOF Operator.

Render Elements:


RPF:


Both images were done with Vray 1.5 RC2. AA Filter : Area. Adaptive Subdivision: 1 2.