Recording 16 bit per channel

K

Kosher

Member
Is there any way to record to computer at 16 bits per channel? How do the film pros take the 10bit YUV encoded video and convert it to HDRI?

I have the A1 and I'd like to try this out. Is there any way to take the straight output from the cam and go right to the computer as RAW format 24P?

Thanks!:love4:
 
That's kind of what I was thinking. Is this the only workflow that I can use? Is there a card like this for a laptop or an external box to put the card in so it works with a laptop?

Thanks!
 
You won't get much more than you'll get on tape.

Component out is 1440x1080x8 with 4:2:2 sampling pre-mpeg-compression, BUT it is subject to analogue noise and resolution loss, which may lose you more than you gain. Tests show that the extra cost and hassle isn't worth it unless you're doing shitloads of greenscreen OR super fast motion stuff which mpeg2 codec falls over.
 
epicedium: What you gain in color information gets you closer to HDRI. I am going for a stylized look that can't be achieved with 8 bits per channel. The max I can get is 10 bit? This is straight analog though so isn't there a way to get 12bit if I capture component straight to the laptop?

Can you point me to any "tests" that you are referring to?

Thanks!
 
Kosher said:
epicedium: What you gain in color information gets you closer to HDRI. I am going for a stylized look that can't be achieved with 8 bits per channel. The max I can get is 10 bit? This is straight analog though so isn't there a way to get 12bit if I capture component straight to the laptop?

Can you point me to any "tests" that you are referring to?

Thanks!
Click to expand...

Elton, one of the members on here did a test a while ago using the component out for the A1. He posted grabs of the uncompressed version vs. normal footage shot to tape. There's a (very) slight difference. You can do a search on his posts to find that thread.
 
HDRI is a bit of a buzz word... My background is in 3d vfx (8yrs) where the day to day realities lie somewhere between 'hdr for the sake of it' and the age-old high bit depth rendering to give more lattitude for re-toning further down the pipeline...

I don't see the relevance, what are you trying to achieve? Are you talking about blending of multi-exposure photos, etc.? You're shooting on a peice of equipment which is at 'least' a generation or two away from a video/HDR type workflow. I suppose that you could conceivably shoot two cameras side by side, if the subject was far enough away to negate the difference in perspective.

But, back to the main point-- considering that you'll get 7 stops capture lattitude if you're lucky- what exactly are you going to do with those 12+ captured bits per pixel? You'll get some amazing expansion of the middle of those 7 stops, but no doubt that's not what you're interested in-- you'll find the top and bottom end just as clamped as if you'd captured quantised to 8 bits.
 
Last edited:
Thank you for the info. I think you know the objective and you provided the answer I was looking for. The hardware is at 'least' a generation or two away from an HDR type workflow. I am not looking for multiple exposures as much as I am the bit depth for lighting, which is something that I can only find available on photo cameras today. The fastest photo camera I can find is 10FPS.

I come from a software development background and also work in film here in the U.S. This side project is just a bit of a hobby and I am looking at some interesting techniques on generating HDRI from standard footage. This is a bit of what I am looking at:
http://portal.acm.org/citation.cfm?id=882270

There's much more though...

I was just wondering if there was some way to put the CPU load on a laptop to process raw data at maximum bit depth. I guess the same question would apply to the photo camera since it's maxed out at 30 raw photos sequentially.

Thanks for your response.

Cheers,
Kosher

epicedium said:
HDRI is a bit of a buzz word... My background is in 3d vfx (8yrs) where the day to day realities lie somewhere between 'hdr for the sake of it' and the age-old high bit depth rendering to give more lattitude for re-toning further down the pipeline...

I don't see the relevance, what are you trying to achieve? Are you talking about blending of multi-exposure photos, etc.? You're shooting on a peice of equipment which is at 'least' a generation or two away from a video/HDR type workflow. I suppose that you could conceivably shoot two cameras side by side, if the subject was far enough away to negate the difference in perspective.

But, back to the main point-- considering that you'll get 7 stops capture lattitude if you're lucky- what exactly are you going to do with those 12+ captured bits per pixel? You'll get some amazing expansion of the middle of those 7 stops, but no doubt that's not what you're interested in-- you'll find the top and bottom end just as clamped as if you'd captured quantised to 8 bits.
Click to expand...
 
Last edited:
Ah okay, yeah I've seen these, this is one of those uber-complex next generation video/image mapping techniques, much like the technique which dynamically maps (in 3d) a set of photo stills onto live video footage to dramatically increase video resolution... Interesting stuff of course :) But the trick is in the software intelligently mapping pixel/spacial relationships over time, and to say that it's complicated would be an understatement :) You're right to bring this stuff up though, lots of interesting stuff on the horizon. At this time it's hard to say how much of this will be niche/speciality and how much can have a real impact on video production.. my hunch is the latter, as even stuff like re:vision's twixtor doesn't have a real impact on video production in general, it's just used for specialist time remapping for specific shots and the occasional conceptual promo, if you know what I mean.
 
http://bourn2fly.tripod.com/id2.html

Not the best site but it has some interesting information. ILM actually started using HDRI in their pipeline a long time ago and uses it today for most occlusion shading and global illumination matching with live action.

It's certainly expensive but it's within reach here shortly. The paper I sent you actually samples temporally, not spacially. There's a loss of frames for the multiple exposure stuff and then it's tone mapped back (like you said).

Mapping motion over time is part of the issue for sure. They've come up with some ideas on how to solve this using motion blocking in post (sort of like green screening).

The Canon XH A1 is actually not too far off. It allows you to capture stills to the SD card while shooting a video but unfortunetly, it doesn't yet allow the auto bracketing feature while shooting the stills. It also needs to allow for simultanious continuous photos while auto-bracketing.

It might be easier to line up the vectors on a canon photo camera and shoot 10 FPS in RAW then do the temporal / motion matching with the algorithms presented in that whitepaper. Photomatix also has batch processing software for the tone mapping, motion matching and hdr generation. It would just be a matter of running a couple steps in there.

I guess this is just something we'll see in the future. I just wish I knew if there were some way (usb?) to control the exposure on the camera while capturing video. At 60 FPS 1080 (autobracketed), I would probably have enough spacial and temporal data to capture the light information. I just need to write the software to control the canon but that's kind of what I do so I was looking for a bit of Canon hardware API direction :)

Thanks!
 
Last edited:
You must log in or register to reply here.
Back
Top