Jump to content


Photo

"HDR" motion video using 2 cams & a beamsplitter mirror?


  • Please log in to reply
15 replies to this topic

#1 Peter J DeCrescenzo

Peter J DeCrescenzo
  • Sustaining Members
  • 620 posts
  • Camera Operator
  • Portland, OR, USA www.peterdv.com Blog: http://herefortheweather.wordpress.com/

Posted 06 November 2009 - 08:07 PM

File this one under weird ideas:

EDIT: Duh, well I've now done a Google search on "HDR video" after I typed all this, so it looks like I've got some reading to do ... :-)

I'm curious if anyone here has experience with or has heard about shooting "HDR" motion video (not "traditional" HDR digital stills) using two identical video cams mounted at 90-degrees to each other and shooting through a partially-reflective beamsplitter mirror?

Similar to a teleprompter with its half-silvered mirror mounted in front of a single camera, but with a 2nd camera replacing the teleprompter CRT/LCD screen?

Both cams would frame the scene identically & use the same focal length lens, but one cam would be exposed for the shadows up through to the lower mid-tones, and the other cam would be exposed for the mid-tones up through the highlights.

The resulting video would be composited in post to (maybe?) yield video with a wider dynamic range than the cams would otherwise be capable of individually. For that matter, the resulting dynamic range may well exceed what's possible with an individual high-end camera costing much more than using 2 inexpensive cams instead.

This part makes my head hurt: Would you need to use a luminance-key to properly composite them together?

The following company sells beamsplitter mirrors, including 3" & 4" samples which might be large enough to at least experiment with (starting at only $6 US!):
http://www.stereosco...rors.com/3d.htm

Anyone have 2 identical video cams & lenses they want to experiment with? Alternatively, a proof of concept could probably be done using 2 _still_ cams?

Anyway, I'd appreciate hearing comments & opinions about this, and whether or not something similar has already been done, or if there is a related camera/product/add-on being sold, or maybe this idea should be filed under "more trouble than it's worth"?*

- Peter

* Among other things, I can see how follow-focus might be an interesting, but not insurmountable, challenge.

EDIT: Change thread title & text from "stereoscopic" to "beamsplitter".
  • 0

#2 Chris Millar

Chris Millar
  • Basic Members
  • PipPipPipPip
  • 1642 posts
  • Other

Posted 06 November 2009 - 08:52 PM

The number of people playing around with this at home would probably number in the 1000's by now

Not many people talk about it online as they think they're the only ones and don't want to give the idea away - occasionally they pop up here and make a claim, only to receive a barrage of links to other people with a bunch of footage on youtube claiming a gazillion stops of latitude.

I don't mean to say its not worthy, or interesting - etc... just be ready to have your thunder stolen by individuals and groups with a lot more time, money and/or education under their belts ...

Have fun ! ;)
  • 0

#3 Phil Rhodes

Phil Rhodes
  • Sustaining Members
  • 11944 posts
  • Other

Posted 06 November 2009 - 09:33 PM

You derive the alpha channel for the highlight pass using its inverted luminance.

That is: in Photoshop you put "dim" layer over "bright" layer. You select all and copy "dim". You enter quick mask and paste, then invert. You exit quick mask and hit "delete". If you have fringing problems you can blur the mask a few pixels, at the cost of glows on sharp luminance edges.

If it's anythig like doing HDR with DSLRs, you will then want to put a pronounced S-curve on it, which the literati would call a "viewing LUT", then a nonlinear desaturation of anything particularly saturated.


P

Edit: I think the excessive saturation issue may be caused by the way Photoshop does a colour-to-monochrome conversion in the circumstance where you're pasting RGB data into an alpha channel. You might experiment with other approaches, perhaps one based on the 709 Y channel derivation. Computer software tends to use either sum(RGB)/3 or max(RGB); the latter will cause oddities. Or better yet, do it per channel, though that might cause auto-knee-style strangeness where the "dim" channel clips a channel, depending on the relative linearity of your CCDs. Argh, this is complicated.
  • 0

#4 Chris Millar

Chris Millar
  • Basic Members
  • PipPipPipPip
  • 1642 posts
  • Other

Posted 06 November 2009 - 10:01 PM

Phil, from what I've read is one of those with lots of - or at least enough - experience (education) under his belt - and to quote him:

Argh, this is complicated.


:lol:
  • 0

#5 Peter J DeCrescenzo

Peter J DeCrescenzo
  • Sustaining Members
  • 620 posts
  • Camera Operator
  • Portland, OR, USA www.peterdv.com Blog: http://herefortheweather.wordpress.com/

Posted 07 November 2009 - 01:26 AM

Hi Chris: I'm not the least bit worried if others have tried or are trying to do something like this. Quite the opposite: I'd like to know if this idea might work, or if it's already "working".

I'm not interested in being an inventor or "productizing" this idea -- I'm just interested in learning if ...

2 identical video cameras mounted at 90 degrees, using identical lenses set at different f-stops*, concurrently shoot a live action scene through a partially-silvered mirror. Then composite the 2 recordings into a single video displaying higher dynamic range than either camera is capable of alone.

... might "work"?

I understood a small fraction of what Phil posted. Maybe. Gives me something to work with. Thanks Phil! :-)

* Oops: Using 2 f-stops would result in each source video having a DOF different compared to the other. Cool! (Not.) So, to avoid that you'd instead set the f-stops identically and adjust exposure on 1 cam using ND filters.

Most of the "HDR videos" examples I've found so far via Google & YouTube are digital still timelapses, or involve shooting a fairly static scene twice with one video cam. That's cool, but not what I'm interested in.

Fun!
  • 0

#6 Chris Millar

Chris Millar
  • Basic Members
  • PipPipPipPip
  • 1642 posts
  • Other

Posted 07 November 2009 - 04:37 AM

Yes it will work.

Better is to use one lens and have a beam splitter behind that going to multiple sensors ... But that is harder to do with limited resources, so in lieu of a R&D dept. try this out:

Build your rig... take a stills camera - shoot a static scene with your camera in position 1, move it without bumping registration to position 2, shoot the scene again ... Take the two stills and stick them in your favorite HDR plugin/prog and well ... you know the rest :P

As an aside - what do you mean by 'stereoscopic' mirror ? I take that as a front surface mirror with a kink in the middle or ?
  • 0

#7 Peter J DeCrescenzo

Peter J DeCrescenzo
  • Sustaining Members
  • 620 posts
  • Camera Operator
  • Portland, OR, USA www.peterdv.com Blog: http://herefortheweather.wordpress.com/

Posted 07 November 2009 - 12:12 PM

As an aside - what do you mean by 'stereoscopic' mirror ? I take that as a front surface mirror with a kink in the middle or ?

Hi Chris: I suppose I should just refer to it as a "beamsplitter". I've edited the title thread & text accordingly.

"Stereoscopic" is the term used by the company of the same name (linked to in my 1st post above). Apparently they make & sell a variety of mirrors.

Your suggestions are appreciated.
  • 0

#8 Phil Rhodes

Phil Rhodes
  • Sustaining Members
  • 11944 posts
  • Other

Posted 07 November 2009 - 01:11 PM

My approach to this was to use one of the JVC high-def cams that does 50p, put an LCD shutter in front synced to dim every other frame, and sort out the result with motion compensated interpolation in post.

P
  • 0

#9 Chris Millar

Chris Millar
  • Basic Members
  • PipPipPipPip
  • 1642 posts
  • Other

Posted 07 November 2009 - 06:27 PM

My approach to this was to use one of the JVC high-def cams that does 50p, put an LCD shutter in front synced to dim every other frame, and sort out the result with motion compensated interpolation in post.

P



As a general rule how does one get the required sync from any given video camera and by whatever means necessary get a nice old fashioned square wave from it ?

...say an EX1 - which I'd have a couple of applications where I'd use that ;)

or did you just have it dimming at 25Hz and get it into phase manually ?
  • 0

#10 Phil Rhodes

Phil Rhodes
  • Sustaining Members
  • 11944 posts
  • Other

Posted 07 November 2009 - 06:34 PM

Oh good grief I never actually did this - whaddayathink I am, a nutcase?

Usually you'd take a composite output and use a sync separator IC to derive a square wave from it. You'd need to modulate that square wave with an AC waveform to drive an LCD shutter panel. It's not rocket science.

cvbs_1line_s.jpg

That's one scanline of composite video (showing colourbars). The big negative-going bit right at the start is the sync pulse.

P
  • 0

#11 Chris Millar

Chris Millar
  • Basic Members
  • PipPipPipPip
  • 1642 posts
  • Other

Posted 07 November 2009 - 08:05 PM

The big negative-going bit right at the start is the sync pulse.


Thats all I need - cheers !

I've realized just now how busy I've been at 'real' work ( :rolleyes: ) to just hook up an o-scope to an output and have a look, I would have seen that dip and used it already - bah!

I've been wanting to sync up strobes for instance - but that was a whiles back and the heat isn't on anymore (we used 16mm instead) - sure I'll come up with something else sooner or later
  • 0

#12 Phil Rhodes

Phil Rhodes
  • Sustaining Members
  • 11944 posts
  • Other

Posted 07 November 2009 - 08:10 PM

Bear in mind that's the -line- sync, not the -frame- sync...
  • 0

#13 Chris Millar

Chris Millar
  • Basic Members
  • PipPipPipPip
  • 1642 posts
  • Other

Posted 07 November 2009 - 09:00 PM

Bear in mind that's the -line- sync, not the -frame- sync...


Ok, I'll admit I'm not a video engineer but I imagine I can use a comparator or similar to get the frequency from that negative pulse and then simply delay the corresponding 'clean' pulse to the position in the cycle that I need ...

Yes, there will be some dicking about ;)
  • 0

#14 Chris Millar

Chris Millar
  • Basic Members
  • PipPipPipPip
  • 1642 posts
  • Other

Posted 07 November 2009 - 09:22 PM

For the OP:

http://www.cinematog...showtopic=41162
  • 0

#15 Ricardo Motta

Ricardo Motta

    New

  • Basic Members
  • Pip
  • 1 posts
  • Cinematographer

Posted 16 November 2009 - 01:52 AM

File this one under weird ideas:

EDIT: Duh, well I've now done a Google search on "HDR video" after I typed all this, so it looks like I've got some reading to do ... :-)

....


It can be done, and it has been done many times. My company has developed and sold near a million HDR video sensors since 2003. www.pixim.com. Francis Coppola was an early supporter as we wanted to develop something for the entertainment imaging, but nobody cared about HDR, and we end-up focusing on CCTV. Still possible if anyone has deep pockets and wants the edge.

Ricardo J. Motta
  • 0

#16 michael abraham

michael abraham
  • Basic Members
  • PipPip
  • 36 posts
  • Other
  • London

Posted 28 November 2009 - 11:11 AM

Hi Peter,


thanks for the useful suggestion you gave us about the company who make stereoscopic mirror.
I was looking from weeks about a company just specialized on them.

Thanks a lot!!!!


Michael
  • 0


Gamma Ray Digital Inc

Abel Cine

Rig Wheels Passport

Paralinx LLC

Opal

The Slider

Ritter Battery

rebotnix Technologies

Wooden Camera

Technodolly

Tai Audio

CineLab

Metropolis Post

Broadcast Solutions Inc

FJS International, LLC

Aerial Filmworks

CineTape

Visual Products

Glidecam

Media Blackout - Custom Cables and AKS

Willys Widgets

Opal

Ritter Battery

Media Blackout - Custom Cables and AKS

The Slider

Paralinx LLC

Wooden Camera

Visual Products

rebotnix Technologies

Metropolis Post

Tai Audio

Rig Wheels Passport

Willys Widgets

Aerial Filmworks

Abel Cine

Gamma Ray Digital Inc

Glidecam

Technodolly

CineTape

CineLab

FJS International, LLC

Broadcast Solutions Inc