Could Red be a 3D camera?

[dillont]

I just had the silliest thought. And then I thought about it. And it doesn't seem so silly after all.

The Mysterium sensor is so large that two 720x480 frame sizes could be potentially be recorded simultaneously on the same chip, right? And given the light patterns falling on the chip from slightly different angles through the lens, wouldn't it be possible in post to marry up the two simultaneously recorded images, and thus creating a pseudo 3D image, or at least an image with a superior depth to standard?

I'm no expert at lenses or cameras by any stretch of the imagination, but I thought I'd posit the possibilities of recording simultaneous images on the chip at the same time. Insipred by the "can I take a still photo while recording a stream?" thread.

Thanks for your time.

Dillon Thomas

[Craig Ryan]

You could theorectically record even FOUR 1080p frame sizes with Red's sensor.
I'd imagine you could do it...It would be an interesting test for sure. It would probably be an intricate process of cropping and creating the exact frames for the 3D effect though.

[Robert Sanders]

I think James Cameron pre-ordered a bunch of RED cameras. I'm sure he has plans to use them on his stereoscopic rig.

I was thinking you could record two 2K images (turned 45 degrees) and split them on the sensor.

[Clint Johnson]

I'm trying to imagine a dual lens system that would be less expensive than the old standby of putting two cameras side by side and everything I come up with is more convoluted, cumbersome and costly. If the Red camera was the usual $120,000 high end camera it might make sense... but at $17,500 it would look to make less sense.

But someone with the time access to a good machine shop might want to play around with a couple Nikon 28mm lenses and a few mirrors. Turning the camera ninety degrees would let you record two 2k images one above the other with at least a little spacial separation for fine tuning the lens mounting system.

[NickJushchyshyn]

Heh heh...
You could always get started by taking a route similar to what the folks that put 35mm adapters on dv cams use today. Just mod an actual stereo photo camera with some moving ground glass into a something that can be mounted onto a RED lens and you're good to go. :)

[Homersapien]

I wouldn't be suprised to see third party manufacturers working on/thinking about doing something for Red like this. High resolution 3d cinema for under $40,000... that'd be funny as... all we would require then is for Red to produce a projector for it that was cheap enough for the average cinema to encorporate!

[AuditoryVisuals]

I think that would be a good idea, but I couldn't see the RED team doing that. Jim seems like he's focused on one thing only.

http://www.dvxuser.com/V6/showthread.php?t=70214 My post about 3D.

[Barry_Green]

I just had the silliest thought. And then I thought about it. And it doesn't seem so silly after all.

The Mysterium sensor is so large that two 720x480 frame sizes could be potentially be recorded simultaneously on the same chip, right? And given the light patterns falling on the chip from slightly different angles through the lens, wouldn't it be possible in post to marry up the two simultaneously recorded images, and thus creating a pseudo 3D image
Not with a regular lens. Yes it'd be possible to actually extract two SD-sized pictures from the sensor, but you'd need to use two lenses mounted "in stereo" in order to image two slightly-different views of the same image.

If you extracted two 720x480 patches off the sensor when only using one lens, the result would be pulling out a section of the left half of the picture, and also a section of the right half of the picture; not two copies of the same image.

So if Red were to design a 3-D lens that projected two images side-by-side on the sensor, then yes it does seem like you could do 3-D in one camera by processing and extracting the images in post.

[Craig Ryan]

Barry, thanks for clearing that up! Makes perfect sense; I had a feeling it had more to it than that.

[stokestack]

You could do it with an image splitter, which would gather a left image and a right image and bounce them into an over/under configuration into the camera. Or just split the image and turn the camera sideways.

I was looking into this for my Canon 20D, but I can't find the link someone sent me. This is an example of the basic idea, although a bit cumbersome because of its orientation:

http://www.ledametrix.com/prism/index.html

The problem with using two cameras (aside from synchronizing their shutters, focus/zoom, and exposure) is the interocular distance. The lenses need to be centered about 64mm apart (the spacing of human eyes); most cameras are too large to sit that close together. You could address that by offsetting the lenses with a periscope-type assembly, however.

This does bring up an interesting topic with applications beyond 3-D: synchronizing Red cameras. I wonder if there'll be any provision for syncing Red cameras not just at the timecode level, but down to the shutter.

[Homersapien]

Thought I'd have a go just for a giggle lol... I know it's very basic but gives an idea :), of course you'd have to split the image in post. Shoot in windowed mode at 2260 x 2540. Two of those lenses is pretty much overkill though lol... Be interesting if the modularity of the camera and rails could be altered to tilt the camera sideways and then an adapter to hold both lenses in place and have a follow focus positioned between the two in order to pull focus on each at the same time... By the way I know it's shit :)

http://www.dvxuser6.com/uploaded/17514/1159907752.jpg

[Homersapien]

Actually that sensor image is slightly wrong... it should be the other way round... one sec I'll change it


http://www.dvxuser6.com/uploaded/17514/1159908227.jpg

[kyle.presley]

I think two cameras would have to be used, IMO. It is my understanding that with 3D cinematography, that you have to set a convergence point for different distances (like your eyes do when you focus on something closer or further away; crosseyed). I believe this convergence is set at the same place the fous is set. Therefore, one would need two separate cameras so that their positions in relationship to each other could be changed. I could be wrong...

[NickJushchyshyn]

Or just stick one of these on the front:
http://www.berezin.com/3d/3d_lens.htm