Plenoptic 3D...

[Evin Grant]

Anyone looking into this as a one lens 3D capture solution looks super cool and configurable...

http://photorumors.com/2010/09/23/wh...noptic-camera/

http://www.raytrix.de/index.php/demo.190.html

[George Tsai]

Cool... tho don't quite understand their graph here...



standard plenoptic camera has 0 effective resolution and standard camera is in the negative when refocusing in post?

any idea how much that camera costs? don't want to send request for the price since i've no interest in actually purchasing, just curious:-P

[Brian Petrie]

Yeah it's neat that they will build you a lens system for any camera; if you want something other than what they offer. Depending how well it works it could be neat little niche option to offer.

[Jim Geduldick]

Looks like some of the fps on the cams is super low at 15 fps and under.
It would be great if Adobe updated us on the tech preview from last year .
Some of the examples on the Raytrix site seems like a bit to much warping of the pixels.

[Seth Larney]

Very cool

[Simen]

Now I can see a practical use for the 28K Mysterium X sensor, as the Plenoptic / 4D Lightfield Technology needs an Ultra Ultra High Resolution sensor to make Ultra High Resolution 3D/Omni-Focus images. ( Are Jim and the other RED guys reading all suggestions? )

Though I do not understand all the practical implications of the graphs, I do understand some of the results - which are in much better resolution and with less artifacts than the demo images I saw a year ago:

Examples of Omni-Focus ability

[http://www.tgeorgiev.net/Gallery/Fredo.gif]
http://www.tgeorgiev.net/CVPR2010/SeagullF.html

I would like this as a second chance to fine adjust the focus and to adjust the depth of field - and wouldn't the areas not in focus usually need less resolution? Is the difference between Standard Plenoptic Cameras and Raytrix Cameras (4D Lightfield Technology) in the software? Can RED make its own softwere, were we can choose from the type of scene where we need the most resolution; closer to the lens focus or spread out all over the scene - or something in between?

The 3D I wont understand, unless the lens diameter in the front is at least as big as the distance between our eyes.

3D-stereo demo

[http://www.tgeorgiev.net/Gallery/Zhengyun1.gif]
http://www.raytrix.de/index.php/id-3d-demo-235.html

The Plenoptic Camera
http://www.raytrix.de/index.php/models.html (4D Lightfield Technology)
http://blog.laptopmag.com/never-take...oto-technology

Explained
http://www.tgeorgiev.net/FocusedPlenoptic.pdf

__________________
Simen Tobiassen
Imagination-Realization-Facilitator

[Dan Hudgins]

For 3D stereo use you only need two views output, and a wider single lens to take with. There is some ease of use to focus with a single large lens, like in the Sony 3D camera, but for shooting 3D its lower cost and lighter weight to just sync two small cameras, like SI-2K mini does.

For multi-view camera if the single lens is smaller than about 125mm wide, when you make a multi-view display you don't get much parallax and head movement.

Many years ago I published some papers about 3D autostereoscopic displays.

In that work was such a concept for a viewer that could display a million views, but it did not need ALL of them at the same time since the viewer's eyes are not able to see all the 1x1mm exit pupils over a 1000x1000mm area. Rather you can see maybe 5x5 for one eye and 5x5 for the other eye, so you only need 50 views out of the million possable. Within the 25 or more for each eye, you get the focus through that a camera such as in the first post gives but under the control of you eye as in natural vision, that is acconnodation depth clue. To only display patches of views head tracking is needed, so the patches that are active are moved around using light valves as needed.

One advance they could use is to get rid of the micro lens array and go to a single imaging array, but a faster one, then the number of views is unlimited if you sequence the entrance pupil with a light valve array, its the same more or less as the 3D display I was working on, you just use the system as a camera rather than a viewer.

To do that two large lenses are required (on the same axis, they act as sort of field lenses and image forming lenses at the same time), as the micro lens array is replaced by another large lens and a light valve array. The upper limit on the number of views is diffraction by the ever smaller entrance pupils, but at one to four million views you need to do less interpolation between them than with the small set of views that the camera takes.

It is interesting that they are able to get good images (see their tech pdf on their site) using so much interpolation, but for 3D stereo use, something optimized for just two spaced views would probably be more useful, like the Sony camera, or better just using a mirror rig with two Epic on it.

The example of the "deep field" camera could maybe be done using the same sensor area with a telecentric lens stopped down, when you shoot many small low noise images, vs. one large high noise image, which one looks better in the end?

That is the same problem as using "anti-blur" processing, what do you gain by using a shutter time that gives you a blured image, the trying to process out the blur, rather than using a fast shutter speed and more noise in the image. The "anti-blur" adds noise also. Does the sharp but high noise short exposure image look better than the "de-blur" long exposure image that is full of processing artifacts. Its non-linear so it can be hard to tell.

Its sort of the same issue with deep-field cameras, if you just stop down you get noise and diffraction blur, so does shooting many lower resolution images that are well exposed then interpolating them give a better high resolution result?

[Matthias Hutter]

It would be nice if you could actually eye-focus on a 3D-TV image - that would reduce a lot of eye-strain.

-Matt

[Dan Hudgins]

It would be nice if you could actually eye-focus on a 3D-TV image - that would reduce a lot of eye-strain.

-Matt

That is one of the things the 3D display system my papers was about, so many years ago.

Its called accommodation depth clue. At that time very few people understood what I was talking about, perhaps, today, maybe more could but one of the papers is "lost" in an obscure procedings.

When you have many exit pupils around the eye you can then focus much like this camera does, you need maybe 25 to 100 exit pupils (per eye), that's why I was involving head tracking and electroptic shutters, to increase the duty cycle so that the display would be brighter and the bandwidth needed would be lower.

Such a system of display can also allow for motion parallax in the horizontal and vertical, it looks much like a white light holographic stereogram (without the rainbows and other hologram issues).

[Simen]

I don't get the point of plenoptic lenses. If you want to change your focus in post, and you're planning this before the shoot, why not use glass that is actually good, with a smaller sensor (like 2/3") and/or smaller aperture to get deep depth of field, then just adjust your focus to any plane you want in After Effects using lens blur and a gradient.

I have seen too much out of focus also on the big screen, I am sure they would have loved a way to have changed it - also choosing a greater depth of field.

When I worked on NEWS spots as a photographer/editor with 2/3" cameras I was well used to automatically know how far to adjust the focus according to how far away I knew the objects were from the camera.

But with my Nikon D90 I am lost when trying to focus with a 50mm lens on my 10 months old daughter when she runs around in the living room (not to mention the focus goes the other way).

I would have loved the plenoptic post focus option in focus critical action situations! Specially if an ultra ultra high res EPIC would still give great ultra high res pictures:)


__________________
Simen Tobiassen
Imagination-Realization-Facilitator