OTOY Forums

I don't think z-depth should be AA.
Having said that, would be cool to have an AA and non AA z-depth for certain tasks.

prodviz wrote:I don't think z-depth should be AA.
Having said that, would be cool to have an AA and non AA z-depth for certain tasks.

It should definitely be optional to do AA. If you use it for example to control a blur mask, i.e. you do depth of field in composite, then the jaggies will show. Also for other stuff.

Vue2Octane wrote:

prodviz wrote:I don't think z-depth should be AA.
Having said that, would be cool to have an AA and non AA z-depth for certain tasks.

It should definitely be optional to do AA. If you use it for example to control a blur mask, i.e. you do depth of field in composite, then the jaggies will show. Also for other stuff.

All our info passes (and our info channels kernels which are the same) do anti-aliasing. You can control it by using pixel filtering but you cannot turn it off. So if you end up with jagged edges it means you did not take enough samples/px.

If there's need for crisp edges, we can make this an option.

cheers,
Thomas

As far as I understand, zdepth represents how far a point is away (in Z) from the camera or image plane. So AA would give the incorrect pixel reading.

So, I'd say we'd want the non AA zdepth first and then a second, AA version (for carrying out other effects.)

Also, on the Passes front, have we sorted out how to output separate red, greed and blue mattes for every material in a scene?
I like the way V-Ray does this by having multiple render layers, each containing red, green and blue.

cheers.

prodviz wrote:As far as I understand, zdepth represents how far a point is away (in Z) from the camera or image plane. So AA would give the incorrect pixel reading.

So, I'd say we'd want the non AA zdepth first and then a second, AA version (for carrying out other effects.)

Also, on the Passes front, have we sorted out how to output separate red, greed and blue mattes for every material in a scene?
I like the way V-Ray does this by having multiple render layers, each containing red, green and blue.

cheers.

Could you have a look at our proposal:
http://render.otoy.com/forum/viewtopic. ... 00#p207200

prodviz wrote:As far as I understand, zdepth represents how far a point is away (in Z) from the camera or image plane. So AA would give the incorrect pixel reading.

So, I'd say we'd want the non AA zdepth first and then a second, AA version (for carrying out other effects.)

Also, on the Passes front, have we sorted out how to output separate red, greed and blue mattes for every material in a scene?
I like the way V-Ray does this by having multiple render layers, each containing red, green and blue.

cheers.

Well not really, if some geometry boundary is visible through a pixel, both are wrong. These are just 2 ways of being wrong.

cheers,
Thomas

stratified wrote:Well not really, if some geometry boundary is visible through a pixel, both are wrong. These are just 2 ways of being wrong.

Yes, exactly, and most of the time an anti-aliased depth pass is much more use than an aliased one.

T.

abstrax wrote:
All three passes together can then be composited onto some background image:

resting_ring_floor_logo.png

With shadows (alpha blending - could be made multiplicative, too):

resting_ring_comp_shadows.png

With reflections (addition):

resting_ring_comp_reflection.png

With the layer (alpha blending):

resting_ring_comp.png

Please let us know if that makes sense to you.

Cheers,
Marcus

Yes, this looks great from a CG/live plate integration point of view. The one thing not covered here though is we would also need to be able to project the bg plate, ie. the image we want to place the CG into, onto geometry that matches it, in this case the floor, and have that geometry be reflected by and pass lighting and shadow contributions onto the CG elements but be invisible to the camera in all the passes. I actually think this would already be possible with the existing visibility options but just wanted to make sure you knew it would be needed in the layer setup as well.

T.

TBFX wrote:Yes, this looks great from a CG/live plate integration point of view. The one thing not covered here though is we would also need to be able to project the bg plate, ie. the image we want to place the CG into, onto geometry that matches it, in this case the floor, and have that geometry be reflected by and pass lighting and shadow contributions onto the CG elements but be invisible to the camera in all the passes. I actually think this would already be possible with the existing visibility options but just wanted to make sure you knew it would be needed in the layer setup as well.

T.

We do that already with the matte material (see http://render.otoy.com/forum/viewtopic. ... 10#p177710) and would do something similar with the layers, at least as an additional option.

hello abstrax.

i went through your proposal again in detail. it still looks fine to me. yet 1 questions remains.

1. concearning reflection pass. what do you mean by "in the image above I also had to add some offset. Otherwise parts of the layer would have negative values which Gimp can't use."


other than that, what TBFX sais about the plate-reflection-comp is very important as well..