[spade111]

Not sure if this will come to generate any interest, but this might be a new direction for development of photographic imaging tools.

Combing the 3d point cloud data as generated by a laser infrared scanning system for z-values (1st generation kinect on Xbox does this to a resolution of 640x480x1cm), with the high resolution output of a quality 2d digital imaging sensor, would allow the computational generation of selective DOF.

You could shoot for example a scene with a few different foreground and background elements, at a narrow aperture (so that many of the elements are within an acceptable dof) and combine this info with depth information from the laser scanner, and select which foreground and background elements in the same scene you would like within a selectable level of focus. Additionally you could use this to artificially simulate (with likely good effect) the out of focus rendering characteristics of many different lenses.

There of course are limitations to this technique.

Here is a generalized proof of concept for scene mapping. All of the elements are in place to confer selective dof:

You Tube

You Tube

---

Last edited by spade111; 03-05-2014 at 10:24 AM.

[Steve.Ledger]

Or buy a Lytro already..

[spade111]

One could indeed go buy a lytro already. Thanks for that.

---------- Post added 03-05-2014 at 02:04 PM ----------

The next step I think is actively being pursued by university (and likely industry) researchers. Fully mapped 3d environments using the point cloud data, with iterative improvements in precision through computational interference interpolation.

You Tube

You Tube

Ostensibly we could use this to generate accurate representations of any lens, any aperture, and any perspective within the scene. Although at this point... We are talking less about photography and more about a virtual environment suitable for consumption via the oculus rift or similar devices.

[spade111]

Here is another video showing a higher resolution iteratively corrected 3d scene. This one is the most impressive I've come across.

You Tube

You Tube

[spade111]

Here is a real time demonstration of microsoft doing iterative precision refinement. From 2011...

You Tube

You Tube

[spade111]

Here is an actual real world use of the technology. Pretty cool Microsoft.
While You Weren't Looking, Dell Made Android Tablets Interesting Again

---------- Post added 11-06-2014 at 01:50 PM ----------

Computational generation of out-of-focus rendering with this technology might finally kill the DSLR.

There is of course limitations to this technology, but it will allow the simulation of any lens. Banking the actual z data from the scene will do much more for the accuracy of the simulation than would artificial techniques via only 2d software.