digital/analog - res/ file size - 3D Cameras/hyper future

[P3D Larry Berlin <lberlin@xxxxxxxxx>]

>Date: Sat, 14 Dec 1996 23:36:39 -0500
>From: P3D John W Roberts writes:
>..........snip (excellent comments)...........................
>............................
> - A note of caution on lossy compression: ......... But 3D viewing is usually
>     much more sensitive to artifacts, especially if the artifacts don't
>     match in the left and right images. I believe that investigation into
>     this issue would be very beneficial in the long term for 3D photography
>     and imaging.

***********  I agree. I just checked the web pages
(http://fi-www.arc.nasa.gov/ic/projects/bayes-group/group/super-res/)
pointed to by Gabriel recently when I mentioned the possibility of obtaining
super resolution and 3D from multiple images. The web pages are great and  I
believe these processes point to useful ways to store massive amounts of
visual data as lower resolution sets, representing the same target image.
The algorithms they currently use obtain a 10:1 improvement of resolution
(10 pixels from each original pixel). Of course they are using 20 some
frames in combination, but they were from roughly the same angle and
lighting so had far fewer differences to work from. As we know with 3D, it's
the differences that count.

On a related  web site (NASA's Millenium Project) they indicate that the
wider the base between the combined source images the higher the resolution
possible from the greater amount of difference information. One of their
projects will establish a triangle of cameras two Kilometers to a side and
they expect to be able to get detailed images of deep space objects that
today's best sees as pinpoints of light. As applied to stereo photography,
this means that two wide based, hyper images could be considered as storing
a higher detailed stereo image than a pair that is ortho. If the hyper pair
could be easily processed by related special algorithms you could derive a
higher-res version which could be computed to show an ortho reconstruction.
This process could theoretically further reduce the stored file size or
increase the resolution from existing file sizes by very significant
amounts. Applied to lossy file reconstruction, it could insure for stereo
imaging that artifacts are matched on each side and are blended in with the
*known* information.

As potentially applied to today's digital camera technology, you could
combine two typical cameras, shoot at accurately measured hyper
arrangements, and process the results for resolution that might match or
exceed today's 35mm film process. I'm theorizing here but it seems that the
more typically available 640 x 480 could amplify sufficiently to this level
of resolution at some ideal hyper arrangement. It might be beneficial to
record at least one image with vertical displacement too for greater accuracy.

Does this mean the demise of '50's cameras? I don't think so at all, though
the convenience of digital is great. It means that eventually you will be
able to digitally reprocess your 35mm film images, that are hyper enough,
into a resolution approaching today's 3D-IMAX formats (and view it as
ortho). When you can readily do that from your home computer, you'll be
quite happy with 35mm and smaller formats, though use of a fixed base stereo
camera will start to take on the semblance of using the single camera
technique. Each stereo sample in a hyper pair would enhance the information
content accordingly so is not a waste. It also means that using the new APS
format will be no problem, shoot hyper. The key here is *multiple images*
from different perspectives, which is already a staple to 3D'rs so we're
ahead of the crowd. 

So, if you're so inclined, bracket your best shots with hypers (currently
requires very accurately measured and recorded locations!) so that next year
you can apply the software tools and get enhanced images. ( When????!!!!,
who really knows but could be soon ..... they are still designing models
that work for multi-spectrum images, ie:color)  : -)

Larry Berlin

Email: lberlin@xxxxxxxxx
http://www.sonic.net/~lberlin/
http://3dzine.simplenet.com/


------------------------------