Issues Tracking or Keying Red Footage?

[MiguelOrtega]

has red team or any directors working with the red tried doing any greenscreen or tracking tests using the red camera?
i know some HD cams have issues with rolling shutters that make it nearly impossible to track in bojou or matcmover etc.

has red conducted any tests on the post production side of the camera?
thanks

[Gavin Greenwalt]

There has been greenscreen footage. It was uncompressed and very nice. However nobody has gotten any good test footage for 3D Tracking. Based on Crossing the line there is definitely a rolling shutter. The impact of that fact has not yet been measured.

There will be an independent test in LA as soon as the first cameras ship in which 3D Tracking and Keying will be a component of the testing. (see #8 Testing thread)

The specifics for the LA RED Test still aren't set in stone. Are there any specific scenarios you would like to see shot?

[Mark L. Pederson]

I am sure that PFTrack 4.0 will track RED footage like butter - we will be testing this as soon as we have our cameras.

[Brook Willard]

I pulled a one-click key on the uncompressed stills they released back in the day... It was the cleanest key I've ever done. Definitely do a search and find those stills to see for yourself. :)

[Don Woods]

Same here. One click on a compressed clip no prob. No tweek looked great. It will be the best digital soulution for keying hands down IMHO. I would beg to say it will be the best option for keying out of all the different types of devices out there.

[Brandon Freeman]

Based on Crossing the line there is definitely a rolling shutter.

Wait, really? That's the same effect that a V1U has when you pan really fast, right? Where the image bends? If RED does that, forget it. :sad:

[Kenn Christenson]

Wait, really? That's the same effect that a V1U has when you pan really fast, right? Where the image bends? If RED does that, forget it. :sad:

The thing is, we don't really know what the actual production footage is going to look like. The Peter Jackson short was shot with prototypes, after all.

I'd definitely wait to pass judgment on this camera until you see footage from an actual "production model."

[Brandon Freeman]

True, but this issue made me nervous about and eventually kept me away from the V1U -- so just the thought of rolling shutter on a $30k production unit makes me jitter. But, yes, I'll have to wait and see. *Fingers crossed*

[Gavin Greenwalt]

Well *every* non-global cmos sensor will have a rolling shutter. Film has a rolling shutter. It's like saying "there is noise in the image." It's obviously there, to what extent is yet to be determined.

[Brook Willard]

Oh, no... Not the rolling shutter debate again :)

[Petros Nousias]

Isnt it possible for the cmos to scan left to right instead of top to bottom so that the effect is more similar to film?

[Brook Willard]

if memory serves, Deanan mentioned that they'd come up with a pretty nice way of rolling the shutter in a film-like way (I.E. what we're used to seeing). That said, i could be completely wrong.

[Gavin Greenwalt]

One could do left to right theoretically but then we would just have a *different* problem.

[Alex Boothby]

Something tells me that the problem would then be 1.78 X more noticeable. :sad:

[Alex Boothby]

BTW: here are some example comps from David Stump's November '06 green screen tests. Start's here:

http://www.dvxuser.com/V6/showthread.php?t=78623&page=2

Blast from the past!

[ChristopherKenworthy]

impossible to track in bojou or matcmover etc.

has red conducted any tests on the post production side of the camera?
thanks

You should try Matchmover on some crappy old DV footage. It works fantastically well. If you use anything better than DV, there's nothing to worry about when it comes to Matchmover. The more info you give it, in terms of focal length and type of move, the faster it goes, but even low-res, badly shot footage can be tracked easily. The Red Footage would be easy. I might give the 1K clip a go later, and see what happens.

[Jeff Kilgroe]

if memory serves, Deanan mentioned that they'd come up with a pretty nice way of rolling the shutter in a film-like way (I.E. what we're used to seeing). That said, i could be completely wrong.

I thought I read something to that effect as well. The rolling shutter issue was definitely acknowledged and we were all told to not judge based on the fact that PJ's short was shot on prototype cameras.

[Jim Arthurs]

Well *every* non-global cmos sensor will have a rolling shutter. Film has a rolling shutter. It's like saying "there is noise in the image." It's obviously there, to what extent is yet to be determined.

About 50% of the film exposure is NOT sheering... it sheers at the beginning and end of the exposure... this does make for a different (and more acceptable) image to track than one that starts slewing at the top and continues all the way to the bottom as the balance of the exposure is more or less sheer proof.

As to the track-ablility of a rolling shutter... well, read this thread by the creator of Syntheyes on his recent attempt to track HV20 footage. He has no axe to grind, and does a good job of explaining how the errors can impact the ultimate tracking result...

http://cybermessageboard.fatcow.com/ssonte/viewtopic.php?t=478

Now in other forums Jim has posted that the RED's shutter is somewhere between the HV20 and film in terms of the sheer, uses a new and different method of triggering, and implies that this is something that could change in future sensor upgrades... please correct me if I'm wrong, or have incorrectly stated.

Should rolling shutter sheer amount be tested as it applies to "track-ability" by the test team? My goodness, it should be the FIRST thing tested with this camera, as it impacts the potential VFX usage in a significant way...

[Michael Schrengohst]

I bought one of those crappy Aiptek GO-HD cameras for my skateborder son.
It has a rolling shutter and makes any action scenes not worth a s**t.
Plus the LCD quit working....Trying to get an RMA. Steer clear.

[Jeff Kilgroe]

Should rolling shutter sheer amount be tested as it applies to "track-ability" by the test team? My goodness, it should be the FIRST thing tested with this camera, as it impacts the potential VFX usage in a significant way...

It will be tested. Tracking and integration are primary on the FX test list and we'll be shooting both indoors and outdoors. Or at least that seems to be the general plan (if what we have could really be called a plan at this point).

[Jannard]

A trained professional (can't say who just yet) did some green screen tests yesterday. We loved what we saw.

Jim

[Jeff Kilgroe]

Thanks for the update/tease, Jim. :)

[MiguelOrtega]

A trained professional (can't say who just yet) did some green screen tests yesterday. We loved what we saw.

Jim

tracking is more important as greenscreen is easier to pull

[Jim Arthurs]

It will be tested. Tracking and integration are primary on the FX test list and we'll be shooting both indoors and outdoors. Or at least that seems to be the general plan (if what we have could really be called a plan at this point).

A large part of the value of RED for my particular needs rests in being able to do 3D tracking and bluescreen/greenscreen. We've know for many, many months how good the RED will key and here I have ultimate faith. The tracking issue is still an unknown and I admit to a bit of personal trepidation until the details emerge.

I have my fingers crossed, my toes crossed, and anything else I can cross with great hope.

I look forward to the test results from all you 3D tracking guru's on the team.

[Alex Boothby]

Jim, have you seen this?

http://prolost.blogspot.com/2007/05/hv20s-rolling-shutter.html

I might try tracking the shot Stu posted under "I've uploaded some footage"...

[Michael Schrengohst]

A large part of the value of RED for my particular needs rests in being able to do 3D tracking and bluescreen/greenscreen. We've know for many, many months how good the RED will key and here I have ultimate faith. The tracking issue is still an unknown and I admit to a bit of personal trepidation until the details emerge.

I have my fingers crossed, my toes crossed, and anything else I can cross with great hope.

I look forward to the test results from all you 3D tracking guru's on the team.

Yes, the sooner we know the better. I have a mo-cap studio that wants
to shoot a feature with the RED. I have a meeting next week. These guys
have already been there and done that with TV and feature credits. It's fun being a tease!

[Brook Willard]

A trained professional (can't say who just yet) did some green screen tests yesterday. We loved what we saw.

Jim

Oh, pffshh :)

[ChristopherKenworthy]

I look forward to the test results from all you 3D tracking guru's on the team.

I was hoping to do a few tests and post them but today's the day I chose to install Adobe CS3 Production Premium, and it's so nervous about installing itself that it's taken eight hours to get my machine back. Anyway, I eventually did some tracking of the Jackson footage with Matchmover. Even at 1K, it's a one-click solve. Simple, simple, simple.

I was going to do some fancy stuff, and put Balrogs and Nazgul in the background and post it here, but then I was reminded about copyright and decided I have better things to do with my time than annoy Peter Jackson.

Suffice it to say, every test I've done indicates that 3D tracking and keying work perfectly with Red.

[Álex Montoya]

Since RED is capable of 60 fps at 4K that means that the cmos is read in less than 1/60th of a second which should be enough for future tracking.

[Don Woods]

Jim you tease... Very cool

[Priyesh P.]

As far as I've understood a rolling shutter produces results like a film camera, so I think the answer to the thread's question should be clear.

[Jim Arthurs]

As far as I've understood a rolling shutter produces results like a film camera, so I think the answer to the thread's question should be clear.

No, not at all! They are different!!! At least when comparing film to the HV20-ish rolling shutter typle we're coming to know...

As I posted earlier in this thread, a film camera exposure has a period where slew can happen, a period when it can't, then a final period when the shutter closes where it can. Here's an animation I did to show the effect. The animation shows a thin horizontal object falling through the FOV in a single frame's time to exand the example... a worst case scenerio.

http://ftp.datausa.com/imageshoppe/outgoing/SHUTTER_STUDIES/FILM_SHUTTER_SKEW_CLASSIC.mov

In practice, the slew portions will be softer as the shutter is not in the plane of focus, and different shutter designs have different angles of attact to the gate, but they all have the same three phases.

A CMOS rolling shutter similar to the HV20 simply starts triggering running down the image from top to bottom with no period where the image is stable. Anders did a very good animation showing the differences...

http://www.reduser.net/forum/showpost.php?p=47000&postcount=3

Obviously the RED isn't the HV20, but this whole issue is my ONLY concern about the RED at this point, and a concern that 95% of the people using the camera wouldn't (and shouldn't) worry about. However, based on my desired use of the RED, I can't help the fact that it is a concern that I await testing on...