Depth Pass anti-aliased?

Become a member of the CGSociety

Connect, Share, and Learn with our Large Growing CG Art Community. It's Free!

THREAD CLOSED
 
Thread Tools Search this Thread Display Modes
Old 06 June 2013   #31
Originally Posted by Troyan: This raises an interesting question. Or a couple, actually.

First, when using the z-depth info Cinema in OpenEXR or RPF, how is transparency that's less than 100% handled? As you can see in that simple cell test, one of the cells that moves behind another semi-transparent cell is not blurred at all, then is heavily blurred as it comes out from behind. One problem with making a map from Zblur2.

Second, in the case of position pass or z-depth data, why doesn't that just generate a z gradient? Why are there objects at all in the depth maps? If all we're controlling is where the focus point is from the camera and how far from that point forward and back from the focus point things get blurred, why is it necessary to include objects? Shouldn't z-depth data just be plotted points in 3D space? Am I only making sense to me?


The p-pass needs the objects in order to determine where it is in space. Each pixel is assigned not only rgb values but also xyz coordinates. Without any objects in which to generate pixels you would not have that info. But I like the idea of the space itself determining the z-depth though dont know if thats possible..
 
Old 06 June 2013   #32
the objects are needed because a close object occludes the gradient since it is in focus. The only way around that is deep compositing, which AE is a long ways from supporting imo. 2 years at the earliest in my speculation, possibly longer considering how little they support exr to date.

Deep compositing stores, essentially a deepshadow map like used in hair shadow rendering. Unlike your typical shadow map where it is a black and white gradient like a depth map, a deep shadow stores multiple values per pixel, store it can store the front side and back side hits of an object, and even of additional objects further behind. What you get is am image that when you invert, you actually see the other side of the object in the matte. The problem with these is they are huge amounts of data and really nuke is the only compositor supporting it at this point.
__________________
Quote: "Until you do what you believe in, how do you know whether you believe in it or not?" -Leo Tolstoy
Kai Pedersen
 
Old 06 June 2013   #33
Originally Posted by JoelOtron: The p-pass needs the objects in order to determine where it is in space. Each pixel is assigned not only rgb values but also xyz coordinates. Without any objects in which to generate pixels you would not have that info. But I like the idea of the space itself determining the z-depth though dont know if thats possible..


Ah. Yeah, I would think it would be possible to attach a null to a camera and use that as your focus object where it could plot a simple z map from there. I'm talking out of my arse, of course, since I have no idea how this works technically.
__________________
2014 Reel
Company website
Behance Portfolio
HyperactiveVR
I reject your reality and substitute my own
 
Old 06 June 2013   #34
Originally Posted by LucentDreams: the objects are needed because a close object occludes the gradient since it is in focus. The only way around that is deep compositing, which AE is a long ways from supporting imo. 2 years at the earliest in my speculation, possibly longer considering how little they support exr to date.

Deep compositing stores, essentially a deepshadow map like used in hair shadow rendering. Unlike your typical shadow map where it is a black and white gradient like a depth map, a deep shadow stores multiple values per pixel, store it can store the front side and back side hits of an object, and even of additional objects further behind. What you get is am image that when you invert, you actually see the other side of the object in the matte. The problem with these is they are huge amounts of data and really nuke is the only compositor supporting it at this point.


Interesting. Can Cinema generate this info in their EXR maps?
__________________
2014 Reel
Company website
Behance Portfolio
HyperactiveVR
I reject your reality and substitute my own
 
Old 06 June 2013   #35
Nope, Cinema doesn't support Deep Compositing. There are only a few apps, that support it atm. PRman and I think Houdini has DC support too. ANd you need Nuke to work with the Files.

Maybe if they integrate OpenEXR2.0 some day. I think it supports Deep Compositing.

EDIT:

If you guys are wondering, how the values and the scale in the Position Pass in camera world are related, maybe I can explain it to you. If you leave the scale at 1 and your units are set to cm. The values in the B-Channel will represent the distance from the Camera to this pixel in cm.
So if the pixel has a value of 1000, it is 1000 cm away from the camera.

Last edited by Tyrus86 : 06 June 2013 at 07:32 PM.
 
Old 06 June 2013   #36
Originally Posted by Troyan: Interesting. Can Cinema generate this info in their EXR maps?


No the app would first have to support exr 2.0 and then add in the deep channel support on top of that. So no they don't currently and judging by maxon's previous development cycles (I'm not a tester anymore so please take this with a grain of salt) I doubt that R15 would have it seeing a show EXR2.0 came out in April, by that point Maxon's usually past new features for next immediate release.

Perhaps in an online update, more likely something like R16 if they see the benefit.



I will say as cool as Deep Compositing can be, the passes are so much larger, requiring massive disk space and typically a notable speed hit to work with, that even when you do have the capability your going to be extremely reserved in using it.

I'd also hope that if Maxon considered implementing it, they'd also consider utilizing DeepShadowMaps, which are the same type of data but used to do a Shadow casting map. The advantage over standard shadow maps is they can handle things like transparency and motionblur, allows them to be utilized even in our physical renderer, which currently renders a raytraced shadow even when shadow map is specified, because shadow maps wouldn't work with many of the features/requirements that physical render has.
__________________
Quote: "Until you do what you believe in, how do you know whether you believe in it or not?" -Leo Tolstoy
Kai Pedersen
 
Old 06 June 2013   #37
you should only ever use position pass with 32 bit btw its useless at 8 and of little use in 16bit. I'd love to show how we used in in lighting on seventh son, but well NDA's and all, but a position pass is an amazing pass to creating all sorts of mattes, combined with object buffers the possibilities to matte something without rerendering or making a roto are crazy, and because it's smooth gradation, you can easily get soft or hard mattes for things.
__________________
Quote: "Until you do what you believe in, how do you know whether you believe in it or not?" -Leo Tolstoy
Kai Pedersen
 
Old 06 June 2013   #38
Thread automatically closed

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.
__________________
CGTalk Policy/Legalities
Note that as CGTalk Members, you agree to the terms and conditions of using this website.
 
Thread Closed share thread



Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off
CGSociety
Society of Digital Artists
www.cgsociety.org

Powered by vBulletin
Copyright 2000 - 2006,
Jelsoft Enterprises Ltd.
Minimize Ads
Forum Jump
Miscellaneous

All times are GMT. The time now is 08:46 AM.


Powered by vBulletin
Copyright ©2000 - 2017, Jelsoft Enterprises Ltd.