By Ren Ng
Whom should I contact with question?
Please contact Ren Ng,
rather than Professors Marc Levoy, Mark Horowitz or Pat Hanrahan.
But please check the FAQ first! We have been a bit overwhelmed
with emails, and it is taking us a while to work through the
backlog. We have been replying to every inquiry individually,
but our responses may be quite delayed.
Are there images that I can use?
All the material on our websites, including the tech report,
powerpoint presentation, movies and images are copyrighted and not for
publication at this time. However, we've prepared a single
set of images for publication, so please feel free to use these as
you like! If you would like to credit the photos, I photographed
them all. I only request that you let us know if you have used
the photos, and to send us a copy of the publication that it appears
in.
The images in the archive illustrate digitally refocusing at different
depths, digitally extending the depth of field, and there are slo a
few photographs of the prototype camera itself.
Please note that you may need to download software to unzip the
archive, such as www.winzip.com.
This is a camera technology that lets you choose what's in focus after
taking the picture. Of course normally you have to choose what to focus on before taking the picture, and there is no way to change that afterwards.
Are you taking a 16MP camera and producing roughly
300x300 final images?
Yes, the resolution of the final images is equal to the resolution of
the microlens array, which is just under 300x300 in the prototype that
we built.
We could have equally chosen to use a 1MP microlens array, and
produced 1MP final output images. However, we would not have
been able to refocus those 1MP images as much as we can refocus our
300x300 images. In this first prototype, we chose to maximize
the ability to refocus, because we were interested in studying what
was new about this approach, and we wanted to produce clear examples
of the benefits to digital refocusing. We basically chose
300x300 as the lowest tolerable final image resolution for our
research purposes.
How much one can
refocus with this approach is relatively simple. In simplest
terms, if a light field camera has NxN pixels under each microlens,
one can ideally make any object in an out put photography N times
sharper (reduce its circle of confusion by a factor of N). In
the prototype that we built, we have measured that we can make objects
about 6 times sharper, although if everything were perfect we might
hope to make things 12 times sharper.