Re: [Gimp-user] Light Field Camera

[Burnie West <west ieee org>]

On 10/29/2011 10:54 AM, Frank Gore wrote:
> On Sat, Oct 29, 2011 at 1:47 PM, Johan Vromans<jvromans squirrel nl>  wrote:
> > But what would be required to process the 'raw' (multi-dimensional) data
> > from an LFE camera? Each pixel or group of pixels gets a Z-component
> > that makes the object that this group represents in focus in a
> > particular plane only.
> It currently requires custom proprietary software from the
> manufacturer, no other software can decode those images yet. As for
> the mechanics of how it could be done and how to add support for it,
> that would be best discussed on the dev mailing list. However, since
> there are currently no publicly available output files from those
> cameras, the point is moot.
>
> --
> Frank Gore
> www.ProjectPontiac.com
> _______________________________________________
> gimp-user-list mailing list
> gimp-user-list gnome org
> http://mail.gnome.org/mailman/listinfo/gimp-user-list
It appears (from the underlying thesis) that an array of individual images with 
different
focuses are captured. The thesis itself indicates 30 images located in a square 
array,
with the target image captured in each element. Since the images are distributed 
across
the entire photosensitive array, a fair amount of chromatic dispersion can be 
detected.

For any image, therefore, 30 times as many pixels (more or less, depending on the
subsequent manufacturing details) must be captured. The underlying software has an
apparently complex but straightforward sorting problem, allowing to choose whichever
pixel is at the best focus for the selected image point in the composite.

It appears at first glance that the images in the light field camera picture 
gallery have
at least four focal planes available (https://www.lytro.com/living-pictures/282).

From the image editing standpoint (if this is the case), it would not be 
sharply different
from the refocusing opportunity by providing four successive shots with 
different focal
length. Of course, the chromatic aberration correction would not be available 
this way,
nor would moving image capture be handled.

So the thirty images would necessarily have to be divided into the appropriate 
planes,
presumably using radial symmetry for the chromatic correction.

Seems to me the light field camera is a pretty good idea, and as it evolves over 
time I
suspect it will be quite interesting.

The thesis itself is available from the lytro.com website - 
lytro.com/renng-thesis.pdf

  -- Burnie