For those of you who think the 14 bit color depth on the new
generation of camera bodies is the cat's meow, here's a challenge:
show us that the extra two bits are worthwhile, in terms of smoother
tonal gradients, ability to withstand editing manipulations, etc.
Here's the setup: I downloaded one of the 1D3 raw files to be found at
http://www.imaging-resource.com/PROD...K3/E1DMK3A.HTM
specifically the sample raw file
http://www.imaging-resource.com/PROD.../Y0B0B0494.CR2
will be used here. You can find a full resolution jpeg on the above
webpage (just click on the thumbnail) if you want to see what it
would look like after a typical raw conversion.
I took the raw file into IRIS, a freeware image analysis tool used in
astrophotography. One file was simply bayer interpolated from the
raw using IRIS (under the hood I believe the engine is dcraw), and so
has full 14-bit resolution. Another copy of the file had its raw
values divided by 4 and then multiplied by 4 to truncate the last two
bits (since when dividing by four in integer arithmetic the
fractional part is dropped when dividing by four, the last two bits
are set to zero when scaling back up again by four). While this is
not the optimal way to truncate the two least significant bits, it
will do for the present purpose. Here are the two files:
http://theory.uchicago.edu/~ejm/pix/...s/sample-A.tif
http://theory.uchicago.edu/~ejm/pix/...s/sample-B.tif
Please note that these are
60MB, 16-bit tiff files, so be sure you
want them before trying to download (I hope my server is up to it).
Post-process as much as you like -- stretch the histogram wildly to
push shadows six stops, for instance, so that any issues will show up
at the 8-bit resolution of our monitors. Or stretch highlights so
that banding of lighter tones should be apparent on our monitors.
There are any number of ways of making the distinction between 12-bit
and 14-bit color depth apparent on a standard computer monitor using
Photoshop, if you are sufficiently creative, AND there is something
to be seen. It's not that you would necessarily take such extreme
measures during an ordinary editing session, rather the issue is
whether there is anything in principle to be gained by adding the two
extra bits.
Your task is to discern which file is the 12-bit file and which is
the 14-bit file, by post-processing each in exactly the same way, and
demonstrating that one stands up better to manipulation. If you can
find a difference, show us the proof and tell us what you did so that
we can reproduce your methods. If no one can find a meaningful
difference, then the extra two bits are in practice superfluous (and
wasteful).
Please note that no setting of black/white points, gamma correction,
white balance, curves correction, etc, have been done to these tiff
files -- only the bayer interpolation has been performed. You'll have
to do manually in Photoshop all these other corrections that raw
converters do. I think that's fairer, so that you have available the
nearly raw data without prior manipulation other than bayer
interpolation (and in one case bit truncation).
--
emil
--
http://theory.uchicago.edu/~ejm/pix/20d/
