First Epic model could shoot 3D by itself

[Jay Gannon]

There's still no depth to the image. Technically what your proposing is impossible.
I would love it not to be, but unfortunatly its not that simple.

[Robert McGee]

Well the idea is based on what the outcome from shooting with an IMAX 3D camera. Now an IMAX 3D camera is not required to capture depth, just two images side to side and on a 65mm negative.

So what about a single camera that captures this with a single lens:

[Jay Gannon]

I still don't understand how this could possibly work, the IMAX 3D records to 2 reels, thus you would need two sensors in the Epic, so that the convergence/parallax can be set to give the 3d effect. How do you plan to do that when there is no parallax in a 2d image as taken with a single sensor. When there is no parallax the image must be recreated in Maya/3d compositor to give a 3d effect. It cannot be done by just shaving the corners and offsetting the image and applying a 3d filter to the final image.

[Robert McGee]

How do you plan to do that when there is no parallax in a 2d image as taken with a single sensor?

Well in that case, we stop looking at the software issue and look at the hardware issue? Now for one is there a such thing as a parallax sensor? Maybe not but that doesn't mean that one can't exist, now in theory it would use my system of adding more horizontal lines but the image would be chopped perfectly and the 2 parallax sensors would take both images, capturing depth as well as left and right eye images at 4K resolution. Now there may be the need for two sensors but the idea is still to aim for just one sensor.

Now with just one or two parallax sensors, you 3D outcome should look like this:


Now Jay, is that what you would want to see from such a one camera system?

http://en.wikipedia.org/wiki/Parallax

[Jay Gannon]

Ok now I'm starting to see where your going now.
It may be possible with a custom lens for the the sensor to essentially be split in half (North South) and record an S3D image onto a single sensor. And then reconstruct as usual. But the lens would have to have some form of dual lens and then take a side by side image and place it top and bottom. I believe Sony and Panasonic have been working on something like this already.

But at that stage I think your up to the price of a dual camera system anyway.

[Bruce Allen]

So yes, it is possible for a single camera to pick up two 2D images and get a proper 3D effect. I know it's pretty tricky but since this is a digital camera, there would be a software issue and if that issue isn't resolved and they put it in there, you'll get a 3D image that has very little or no depth at all.
http://en.wikipedia.org/wiki/IMAX#IMAX_3D

No. It is not a software issue.

It is a physical hardware issue - you need to get two images projected FROM DIFFERENT POINTS OF VIEW.

Not two images from the same point of view.

For that you need a clever stereo lens system - of which there are many already and I'm sure there will be ones introduced in the future.

Not a "parallax sensor." Not "adding more lines"

Eg like this:
http://www.loreo.com/pages/cameras/camera_3dcap_1.html
Except add an anamorphic squeeze as well.

Or to use the IMAX 3D camera example, this:
http://viperfs.com/VFXTLK/imax.JPG

To me that was always the point of the largest RED sensor sizes - they obviously were designed to for "medium format stereo".

But the same procedure works on a FF35 or S35 sensor as well.

Now an IMAX 3D camera is not required to capture depth, just two images side to side and on a 65mm negative.

No. you need left-right difference. Eg parallax. Stuff that provides depth information.

Bruce Allen
www.boacinema.com

[Sebastian Gross]

i'd like to add a little something, although i am not a 3D expert, too.

i couldn't get my head round the question how to pull the right eye information from the material if it's just not there... rotoscoping? *yawn*
adding a 2nd camera in after effects? nope. that's just like someone flicks your eyes.
panning a duplicated layer in after effects + adding a 2nd camera? wait a minute. that's just as if i did nothing at all.

so what i tried in an experiment was:

1. perform a stable-as-possible shot h.o.r.i.z.o.n.t.a.l.l.y.
2. put the shot in an after effects comp.
3. duplicate the layer and move the uppermost 1 or 2 frames to the right, so it gives you the "right eye" information, pulling it from 1 or 2 frames beyond... *tumbleweed*
4. apply the ae_onboard "3D-glasses" filter to the lowermost layer, turn the uppermost off.
5. assign the lowermost layer to the left eye, the uppermost to the right eye
6. render in anaglyph.

i am well aware of the fact that this is far from real 3D and it doesn't work with vertical camera moves but if well performed the result is quite "...".

notice(s):
- if our camera move is shaky this method will crush your eyeballs and so will pushing it more than 3 frames. (looking at a freeze frame instead with anyglyph glasses it'll gain more depth the further you push your uppermost layer)

now i am just waiting for comment like: "nah... did that years ago.."

[Robert McGee]

Sebastian, I can tell you and everyone else why my theory on the system holds water.

The sensor in the RED Epic XS35 is capable of 5K, that is 5120x2700. Part of that sensor would be needed to capture two 4K images.

This is the example of the XS35 sensor on the RED website:
http://epic.red.com/epic.html

This is my example of that same sensor capturing 2 4K images for 3D:

As you can see, the XS35 would capture more then what would be required for a true S3D image, and you don't even need no fancy lens if fits on the camera, then it's usable, but if this system is put in to use some people would make special lenses to help achieve a higher level of depth. This is just an example of someone shooting at 1.78:1 at 4K resolution in 3D. As you can see the sensor doesn't add more horizontal lines, those extra horizontal lines are already there, the reason why they are red and blue was to use anaglyph colors to demo this for you. The full range of the sensor makes the images too far apart for a pair of eyes to line them up with 3D glasses so I just went ahead and stretched them as far as they would go anyway. It's actually a full size 5K image made in Microsoft Digital Imaging Standard 2006, the 4K images are also full size. So where is the full size 5K size .jpg, link only:
http://img526.imageshack.us/img526/9...3dexample3.jpg

[Charles Angus]

I think it's officially time to stop feeding the trolls.

[Kyle Presley]

Nikon mount. One of the few ways I could think of to do it with one camera.