ZPass without AA 2004-08-18 - By Luc-Eric Rousseau
Back
this is because MR takes samples at the corder of pixels,
and runs filter in the middles
So with min and max sampling equal to 0,
and box 1,1, you're averaging 4 samples for one result.
I suspect that Triangle 1,1 disables filtering.
> -- --Original Message-- --
> From: owner-xsi@(protected)
> [mailto:owner-xsi@(protected)]On Behalf
> Of Guillaume Laforge
> Posted At: Wednesday, August 18, 2004 6:37 AM
> Posted To: xsi
> Conversation: Things I would like to see in the FX Tree...
> Subject: RE: Things I would like to see in the FX Tree...
>
>
> I tried Zpass without antialiasing too and it seems that boxe
> filtering blur
> the edges even with 0 min/max sampling.
>
> I find that the Lancszos filtering doesn't blur the edges
> when you used 0
> min/max sampling and the X and Y size filtering at the minimum.
>
> Mes 2 euros...
>
>
>
> Guillaume Laforge
> freelance animator
>
>
>
> -- --Message d'origine-- --
> De : owner-xsi@(protected)
> [mailto:owner-xsi@(protected)]De la part
> de Aloys Baillet
> Envoy� : mercredi 18 ao�t 2004 11:53
> � : XSI@(protected)
> Objet : Re: Things I would like to see in the FX Tree...
>
>
> Hello,
> While we are on the AA subject, I tried some time ago to
> compute a ZPass
> without any AntiAliasing. For that I used min and max sampling at 0 to
> only have one sample per pixel, I believed it could give me an image
> without AA. But anyway, MentalRay seems to interpolate the image and I
> had smooth edges.
> Has anybody an idea about how to get pure Aliased images from
> Mental Ray?
> Thanks,
>
> --
> Aloys Baillet - R&D @ La Maison
> --
>
>
>
> Axel Akesson wrote:
>
> >I agree with Morten about this.
> >When I render a depth pass (standard image, no zpic) with good AA I
> >still get artifacts in the edges. This is something you
> don't get with
> >Lenscare.
> >
> >However I take a look at your suggestion, Oscar. Thanx!
> >
> >Axel
> >Virtual Republic
> >
> >-- --Urspr�ngliche Nachricht-- --
> >Von: owner-xsi@(protected)
> [mailto:owner-xsi@(protected)] Im Auftrag
> >von Oscar Ju�rez
> >Bereitgestellt: Mittwoch, 18. August 2004 06:37
> >Bereitgestellt in: XSI
> >Unterhaltung: Things I would like to see in the FX Tree...
> >Betreff: Re: Things I would like to see in the FX Tree...
> >
> >about AA in depth images
> >i dont know about zpics since i havent used them so far
> >i always render in a pass
> >
> >But a depth image with AA can be problematic since the edges of an AA
> >image can produce several gray values, causing a bad info on
> that areas.
> >
> >Frischluft [Lenscare] recomends for their plugin at least, rendering
> >double
> >resolution
> >without AA and using it that way bringing it down to normal
> resolution.
> >
> >
> >-- -- Original Message -- --
> >From: "Luc-Eric Rousseau" <lucer@(protected)>
> >To: <XSI@(protected)>
> >Sent: Tuesday, August 17, 2004 8:42 PM
> >Subject: RE: Things I would like to see in the FX Tree...
> >
> >
> >
> >
> >>The depth of field filter in the fxtree is the same as the Output
> >>
> >>
> >shader
> >of the same name.
> >
> >
> >>It works by calculating the circle of confusion (COC) of each pixel,
> >>
> >>
> >given
> >what's found in the ZPic that's in the second input. Then, for each
> >pixel
> >on screen in cumulates aaalllll the pixels of the image
> weighted by the
> >radius of the COC. It's a classic (another word for 'brute force', I
> >guess
> >:D) implementation described somewhere in a siggraph paper. It gives
> >pretty
> >good result, but I don't know how the ZPics are with
> particles. (BTW,
> >lack
> >of antialiasing of ZPic is pretty much irrelevant in this effect)
> >
> >
> >>What someone needs to do is implement a ZBlur like in shake, which
> >>
> >>
> >isn't
> >technically better, it's just an approximation, but it's
> damn fast. A
> >compound blur is also something we need. Amazingly, afaik we have
> >neither in
> >any product at Avid.
> >
> >
> >>I'll take a look about your turbulant gradient problem.
> >>
> >>In the V4.0 Transform 3D, the default for center of rotation is 0,0,
> >>
> >>
> >in
> >which case it automatically calculates the center of the
> image, so there
> >is
> >no need to adjust it. As soon as you touch it, it is set in pixel.
> >This is
> >because there is a feature in Illusion which is pretty cool, but we
> >don't
> >have in XSI. In Illusion (or on the Avid, now that I think
> of it), the
> >software knows if the user has touched a parameter or not.
> If the user
> >hasn't touched it, it can do a default behaviour. As soon as you
> >touched it
> >then what the user chooses becomes the set value. This is also the
> >magic
> >that makes Illusion always have proper 'default' rectangles. On an
> >Avid,
> >it's even cooler, you can switch between the 'user set' value and the
> >default by clicking on a divot (which has turned blue as soon as you
> >touched
> >a slider). Unfortunately I can't convince anyone else here
> that it is
> >cool.
> >:D It also allows to more clearly skip large parts of code
> when the user
> >hasn't touched paramete!
> >
> >
> >> rs, for example in CC/Keyer, etc. Eventually I want to replace the
> >>
> >>
> >Transform3D node(or maybe make it not matter in another way), but up
> >until
> >now something critical always showed up first.
> >
> >
> >>With regards to centering images in the fxviewer, it isn't possible
> >>
> >>
> >presently. It'll be implemented when we get to add Safe Areas.
> >
> >
> >>>-- --Original Message-- --
> >>>From: owner-xsi@(protected)
> >>>[mailto:owner-xsi@(protected)]On Behalf
> >>>Of Morten Bartholdy
> >>>Posted At: Tuesday, August 17, 2004 6:40 PM
> >>>Posted To: xsi
> >>>Conversation: Things I would like to see in the FX Tree...
> >>>Subject: Re: Things I would like to see in the FX Tree...
> >>>
> >>>
> >>>Thorough and to the point as usual Luc-Eric!
> >>>
> >>>Being an old Illusion user I enjoy using FX Tree more and
> >>>more for each
> >>>update.
> >>>
> >>>I am working on a project where I need DOF to be done in post
> >>>(particles w
> >>>raytraced DOF at HD renders at more than 2 hours per frame!! App.
> >>>
> >>>
> >2,5
> >
> >
> >>>minutes per frame wo DOF and the same for the Depth Capture)
> >>>so I had a look
> >>>at the DepthOfField node in FX Tree. Results were very
> >>>disappointing - no
> >>>nice DOF circles and terrible crawling artifacts when
> >>>foreground particles
> >>>passes background ones. Actually I had not had my hopes up
> >>>very high on this
> >>>one, since I am dealing with tiny spots being DOF blurred a
> >>>lot. I suspect
> >>>it works like creating a selection based on luminance, then blurs
> >>>
> >>>
> >the
> >
> >
> >>>content of that selection using a gaussian filter. It would
> >>>require some
> >>>very clever algorithms to properly blur a foreground
> particle partly
> >>>occluding a background one, especially if they pass each
> >>>other, and I guess
> >>>the FX Tree implementation doesn't take it that far.
> >>>
> >>>I went searching the web and found this nice little AE
> >>>compatible plugin
> >>>which I could give a spin in Combustion. It does a very nice
> >>>job - not quite
> >>>like raytraced DOF, but still. Lenscare from
> >>>
> >>>
> >http://www.frischluft.com
> >
> >
> >>>provides a very nice DOF with aperture control and other neat DOF
> >>>enhancements, speed and high quality. It renders DOF on my HD
> >>>particles in
> >>>app. 1 minute per frame!! Yes, a bit long time for a postfilter but
> >>>considering the raytracing alternative I can wait.
> >>>
> >>>This is not a sales pitch - I just wanted to bring it to your
> >>>attention -
> >>>perhaps it could inspire future updates of FX Tree :) -and a
> >>>user tip for
> >>>those who cant wait for the plugin to reach XSI ;p
> >>>
> >>>On the stability issue - I have experienced crashes in the
> >>>Gradients node
> >>>when turning Turbulence on...
> >>>
> >>>BTW is there a way to make the wiever center the image instead of
> >>>positioning at lower left corner?
> >>>
> >>>Other gripes:
> >>>
> >>>- Would be very nice to be able to import multiple sequences
> >>>in one go, when
> >>>they are in the same directory. (Illusion did that)
> >>>
> >>>- I really don't get the idea behind the default source
> >>>Rectangle in the
> >>>Transform3D node. I'd expect it to default to the input size,
> >>>and center the
> >>>rotation center. Widgets would be nice here too. On a side
> note, the
> >>>Illusion 3D Transform also had its problems. I'd say that
> >>>post sw inside a
> >>>3D app. ought to have sensible 3D transform controls, no?
> >>>
> >>>Enough ranting for now ;)
> >>>
> >>>
> >>>
> >>>- Morten
> >>>
> >>>
> >>---
> >>Unsubscribe? Mail Majordomo@(protected) with the following text in
> >>
> >>
> >body:
> >
> >
> >>unsubscribe xsi
> >>
> >>
> >>
> >
> >---
> >Unsubscribe? Mail Majordomo@(protected) with the following text in
> >body:
> >unsubscribe xsi
> >
> >
> >---
> >Unsubscribe? Mail Majordomo@(protected) with the following
> text in body:
> >unsubscribe xsi
> >
> >
> >
> >
> >
>
> ---
> Unsubscribe? Mail Majordomo@(protected) with the following
> text in body:
> unsubscribe xsi
>
>
> ---
> Unsubscribe? Mail Majordomo@(protected) with the following
> text in body:
> unsubscribe xsi
>
---
Unsubscribe? Mail Majordomo@(protected) with the following text in body:
unsubscribe xsi
|
|
