View Full Version : Whip Pan Artifacts
Jeff Brue
09-08-2007, 09:19 PM
So I've been running through the footage from the test shoot we did last week, and for the most part I'm falling a little bit in love with the imagery in post....1 click greens screens, low noise, good range... The only thing I've noticed so far is this below. Now I understand the math behind a bayer pattern sensor and I've developed my own tricks for dealing with this artifact off of Phantoms and our Medula's, what I'm wondering is can we expect to be able to flag high motion scenes in RedCine to possiblly look at more surrounding pixel groups to level off these lines ? The big problem area is particuarly the orange/yellow head light.
Whip_Pan_Image (http://www.thecamerahouse.com/red_images/whip_pan.jpg)
Kevin Halverson
09-08-2007, 09:24 PM
Its really hard to tell from a single frame grab how this is going to read in motion. Can you post a link to a few seconds of this shot?
Jeff Brue
09-08-2007, 09:34 PM
As with anything getting a couple seconds of 4k or 2k material up is rather painful. If someone wants to host they're more than welcome I have it at a 4k tiff sequence right now and will put it to whatever. In regards to seeing it in motion, it'll be a problem on a lot of TV sets as it looks suspiciously like 3:2 and a lot of consumer Plasma's and LCDs process to remove it (again personal experiance with other CMOS cameras that have the same issue).
Matt Uhry
09-08-2007, 09:36 PM
That does not look necessarily abnormal to me, parts of the car's reflector are very bright, hence the stripes. It's hard to tell from this frame. Any more examples? does this look at all odd in motion?
Matt Uhry
www.mattuhry.com
Poi Boy
09-08-2007, 09:42 PM
I'll bet it looks just fine in motion.
Aloha
-A
Seung Han
09-08-2007, 09:42 PM
can you upload two or three sequential uncompressed files? Also some information about how this was shot...
jbeale
09-08-2007, 09:51 PM
I assume we're discussing the horizontal orange stripes on the turn signal? Is that really a camera artifact? What does that area look like when the camera is not in motion? For example, on many cars, the yellow plastic area over the turn signal has a ridged texture to spread light around, so I would expect to see this sort of thing regardless of camera. But if this particular area is smooth when not panning, then I assume it is the camera.
Edit: zooming into it in Photoshop shows the stripes are slightly more than 6 pixels apart. That is not a bayer artifact, it's too large. It seems mostly likely the real-life pattern of the turn signal light, turned into horizontal lines by the pan motion.
Mike Prevette
09-08-2007, 09:56 PM
Looks fine to me. It's exactly as a blown highlight would read on film.
_mike
kmikami
09-08-2007, 10:06 PM
No rolling shutter skew is visible! That's promising.
I have been closely looking at all the footage so far as well trying to spot wip pan artifacts as they can wreak havoc with auto 3D tracking software. So far I have seen nothing out of the ordinary and I think that this image shows the same. I think what you are seeing is the actual shape of the Jeeps turn signals not a whip pan artifact. If it were an artifact other areas of the frame would show it... like the antenna, or the license plate, tire, ect.
http://img507.imageshack.us/img507/3126/whippan3ry6.jpg
Jannard
09-08-2007, 10:53 PM
Just how fast was the whip-pan?
Jim
James T Mather
09-09-2007, 12:09 AM
It looks perfectly normal to me.
On the upside I see no temporal staggering of the image that I would associate with RS artifacting. Well done to the engineers at RED.
indeed. best CMOS performance I have seen yet.
Jeff Brue
09-09-2007, 02:48 AM
Yeah BK that is just high frequency detail... found a frame with the headlight it has a semi ridge pattern on it. The whip pan was going decently fast Jim, once Rufus got the basketball he wouldn't put it down all day...kind of like the rest of us with the red. I'll be putting this up on our barco tommorow to see how it behaves....
This does bring a question to mind, because resolving that amount of motion blur detail really feels different than scanned film plates... Any ideas on methods to give the motion blur a more organic unresolved look ? I know I know I'm just running over the questions I know I'm going to get asked by DP's on monday.
Álex Montoya
09-09-2007, 03:19 AM
Anyway I would avoid the sharpening in REDAlert! and the low quality JPG's. Far too much macroblocking there.
Ramesh Jai
09-09-2007, 03:42 AM
Maybe that is the whole idea of a whip pan? If you don't want it don't 'whip' pan. I say that will all humility without meaning to be condescending. Thanks.
James T Mather
09-09-2007, 04:18 AM
It seems to me that when people start pouring over freeze frames they can see all sorts of things that are not apparent in a final moving image - However, if it bothers you that much, shoot film.
Jim Arthurs
09-09-2007, 06:42 AM
That's as bad as the RS skew is in a moderately fast pan? I'm SO glad I have a reservation and the only worry I've had about RED since NAB evaporated in one quick image post...
Folks, don't come down on Jeff so hard for simply asking about something, us VFX kinda folks look at things differently, and it's our job to do so. Understanding what we're looking at (so we can manipulate it) it is as important to us as pure esthetics are to you DP's...
Otherwise, your 100 million dollar feature films will all be "The Sad Sister Circle" chick flicks with no cool dinosaurs and sea monster pirates and such.
Chill out.
Darren Orange
09-09-2007, 08:01 AM
Was this camera on a tripod? Does anyone else have any thoughts on the fact that the lines are..... well basically perfectly horizontal? This frame seems a little odd to me. It almost looks blurred via some post effect and then compressed in a fashion that has decrease chroma information. It just seems too perfect in the blur motion, just my two cents, and there is certainly not 4:4:4 chroma here.
Jiri Bakala
09-09-2007, 08:03 AM
regista, everybody has always discussed everything here and so I don't see a reason why change it now. No need to be defensive or secretive. I also don't believe that Jeff is posting his finds as a 'defect'. Red is monitoring these boards closely and they will respond if they feel they need to.
Nobody is crying "Wolf" here. And we all have the right to know if there are any real or perceived problems.
Mercury rising. I like it.
Please, continue,,,
Joe Carney
09-09-2007, 08:19 AM
As someone who is planning on using Red for VFX work, thanks for posting this Jeff. The more info the better.
Ramesh Jai
09-09-2007, 08:41 AM
Of course, the audience will be able to deconstruct every frame of a whip pan with their advanced 1/32000 sec. visual processes.
So let's get the engineers back on it. We must perfect the inherently defective 24fps image acquisition model that audiences have learned to live with for generations so we can get some visually perfect whip pans.
And no, we cannot recreate a perfect whip pan in post, either. That'd be too easy.
Next, the engineers will go out and build a new jetplane, even better than the new Boeing Dreamliner, so it can be flown a few feet over Kitty Hawk for a few seconds at a time.
Unbelievable...
:clown2:
I agree. I think it's much ado about nothing. 99% of your audience don't even know what an artifact is unless your story is so boring that they start looking for every abnormality in the frame.
When I look at all the images that have been taken with RED and posted on this forum, I have to constantly remind myself that the images I am seeing are from a video camera not a still camera. RED is that amazing!
Obin Olson
09-09-2007, 01:54 PM
I agree. I think it's much ado about nothing. 99% of your audience don't even know what an artifact is unless your story is so boring that they start looking for every abnormality in the frame.
Way off base. you don't understand what we are talking about. This is an issue that if it shows in the footage we can't track this camera in 3d space, or the workaround is a real PITA.
Jannard
09-09-2007, 02:07 PM
I agree. I think it's much ado about nothing. 99% of your audience don't even know what an artifact is unless your story is so boring that they start looking for every abnormality in the frame.
Way off base. you don't understand what we are talking about. This is an issue that if it shows in the footage we can't track this camera in 3d space, or the workaround is a real PITA.
I still ask the question as to the speed of the whip pan. We have acknowledged that we are close to a film camera in skew, but we do not match it today (we will in the future). So if you have to slow down the whip pan just a but to get the results you need, this would be valuable info for all.
I talked to Jon Farhat about this and he is good with the performance of the camera for VFX.
Jim
Jeff Brue
09-09-2007, 03:25 PM
Whoa guys...sorry, if i ruffled any feathers. I was by far not trying to come down on the camera. My personal opinion of it is just I'm going to have to explain this stuff to clients who are going to be asking these same questions from vfx supervisors to A list DP's and directors. This stuff gets picked apart on ALL cameras be they Arri 435's or Vipers or F23's. It's what a good rental house does.
After talking to my camera guys the skew isn't an issue as that 4.5 ms thing can be variable as how it eventually shows up on film based on the shutter, lighting conditions, gate weave, and a myriad of other things.
As far as the speed of the whip pan.... We went as fast as we could, and this was the only thing we saw in terms of artifacting, and it turned out to be a non issue.
In regards to contacting Red about possible problems...these cameras are now live and out there the truth comes out. It's fine to be all Apple before a product gets released, but I know my company has 5 reservations, and I know that I'm going to have to give hard answers to clients on what the camera can do, and guess what it can do whip pans ;-)
Ramesh Jai
09-09-2007, 03:30 PM
I agree. I think it's much ado about nothing. 99% of your audience don't even know what an artifact is unless your story is so boring that they start looking for every abnormality in the frame.
Way off base. you don't understand what we are talking about. This is an issue that if it shows in the footage we can't track this camera in 3d space, or the workaround is a real PITA.
Sorry. I got the context. Apologies.
Obin Olson
09-09-2007, 04:24 PM
Looks like it will work fine from what I have seen after the Crossing camera. I assume you have improved this aspect Jim?
Keith Alan Morris
09-09-2007, 04:31 PM
Otherwise, your 100 million dollar feature films will all be "The Sad Sister Circle" chick flicks with no cool dinosaurs and sea monster pirates and such.
Chill out.
lol
thats hilarious.
Jeff Brue
09-09-2007, 04:47 PM
In regards to the images veracity It came from a 4k Tiff out of RedAlert down to a jpeg, with pretty high compression. I was looking to show a possible motion artifact, not the images bit depth...Which for everyones reference I put up there between the Viper and the F23 for amount of latititude. The Red Log format looks like it has a lot of potential. Can't wait to do what we did for the Viper and F23...Mount all three cameras on the hood of a vehicle, and drive around downtown LA and out in the country.
On a side note I'm playing with the underexposed footage from the test in the Digital Vision Film Master today...If anyones having any real worries about noise...It can be gotten rid of fairly easily. Noise Ninja, Furnace, and DV's DVNR tool set provide the best removal in that order.
GlennChan
09-09-2007, 04:51 PM
Er so you're saying that Noise Ninja is #1? Or DVision's?
Jeff Brue
09-09-2007, 05:24 PM
Noise ninja doesn't take into account the motion going on behind the frame so it can sometimes result in artifacting... Furnace from the foundry does a great job its just horribly slow. DV's is about 10-15 times faster (gpu acceleration) and really can pull back in detail that otherwise is lost. Coincidentally each is a lot more expensive in progression.
Jannard
09-09-2007, 06:35 PM
Looks like it will work fine from what I have seen after the Crossing camera. I assume you have improved this aspect Jim?
Vastly...
Jim
Michael Brennan
09-10-2007, 05:19 AM
If a RED user has a concern about something that seems like a pervasive product defect, it'd be best if the issue were discussed with RED technicians directly before posting about it in a public forum. If RED technicians offer no satisfactory answers, or if the defect is verified and their answers and/or solutions would behoove the rest of the RED user population, then by all means, post away.
Things would work so much better that way.
A lot of folks in this forum seem to be overly anxious to be the first to cry "Wolf!" about RED One, it makes me wonder.
:calm:
Ok there will be spurious reports, but keep them public and post frame grabs, this way we learn of re-occuring, spurious non tech issues caused by operator error or post srewups.
For instance if post house says a picture has a particular artifact we would know if it is an effect likely to be caused by a common operator error in say transcoding.
However the main benifit of posting about problems is hearing in a timely manner of real tech isssues and traits.
Sony didn't communicate numerous f900 faults and traits, the f900 community didn't have a forum and since most f900s were owned by rental companies who preferred not to go public there was no body of opinion to demand fixes and so Sony dragged their heels.
As a result DPs around the world were destined to suffer avoidable operator errors and also be subjected to camera faults that undermined confidence in the camera and gave much ammunition to the anti HD fraternity.
Mike Brennan
Gavin Greenwalt
09-10-2007, 10:30 AM
Looks fantastic to me. Over the line was pushing the boundries of becoming annoying but I would sign off on that much skew without concern.
Bang up job Jim. Vast improvements indeed.
Us VFX kinda folks look at things differently, and it's our job to do so. Understanding what we're looking at (so we can manipulate it) it is as important to us as pure esthetics are to you DP's...
Otherwise, your 100 million dollar feature films will all be "The Sad Sister Circle" chick flicks with no cool dinosaurs and sea monster pirates and such.
Vote for sticky. :p
GlennChan
09-10-2007, 12:11 PM
Noise ninja doesn't take into account the motion going on behind the frame so it can sometimes result in artifacting... Furnace from the foundry does a great job its just horribly slow. DV's is about 10-15 times faster (gpu acceleration) and really can pull back in detail that otherwise is lost. Coincidentally each is a lot more expensive in progression.
Thanks for the clarification.
2- I thought Digital Vision's solution was accelerated by custom hardware?
If it were GPU acceleration, then it suggests that some really incredibly stuff could be done on desktop computers.
Jeff Brue
09-10-2007, 11:14 PM
Really cool stuff can be done on desktop hardware... you just need genius's to take it from concept to code (ahem red). As far as the DV toolset they have a realtime hardware box that can do the same thing. However they have a set of plugins for the color grading apps that can do 5 motion estimated frames per second. Versus say Furnaces 5-10 seconds per frame.