John Hermanson wrote:
> I had been using Raw Shooter Essentials for a while but am now trying out Raw
> Therapee. Is there any benefit to saving a processed ORF as a 16 bit tiff
> instead of 8 bit? I don't see it, but 16 bit is twice the file size, 47 meg
> versus 23.5.
>
>
The sensor is an analog device, as are the first part of the sensor
electronics.
The analog to digital converter divides up the analog voltages into
discrete digital steps.
Most DSLRs use 12 bit A/D converters. some high end ones, including the
E-1 and, I assume, E-3, are 14 bit. Extending dynamic range, as some
newer sensor stems are now doing, requires more bits if midrange tonal
detail is not to be compromised.
That translates to 4096 separate brightness step for each color at 12
bits and 16384 for 14 bit.
If you convert that to an 8 bit TIFF, the luminance range is downsampled
to 256 steps, throwing away much of the image data you bought the camera
for. 16 bit's 65536 step range can accommodate all that comes out of
the camera, without loss.
If an image is simply to be converted and viewed as a JPEG (compressed 8
bit) or printed on an 8 bit printer, 8 bit files are fine.
As soon as one starts manipulating the file, even with simple tools like
Levels, Contrast, etc. the software needs to move values around in the
histogram. With only 256 levels to work with, that can lead to uneven
stepping effects, even to holes in the histogram (yup, I've seen 'em),
where working in 16 bit gives the algorithms plenty of steps for subtle
interpolation.
In practice, some images seem to come through considerable manipulation
in 8 bit just fine, while others start to look "funny" in ways that I
find hard to describe. Even starting with an 8 bit image, as in images
from digicams with 8 bit JPEG output only or those downloaded from the
web, conversion to 16 bit before processing can make a difference.
How do I know? I "always" convert 8 to 16 bit in PS as a first step in
editing images from my F30 or Moosterizing images from others. Except
sometimes I forget that first step - and start wondering why things
aren't working quite as I expect. A quick glance at the header leads to
a "DOH" moment and starting over.
Why not 12 and 14 bit versions of TIFF? While certainly possible, such
files would be messier to process. Going from 8 to 16 bit simply uses
two bytes per color per pixel instead of one, whereas intermediate
numbers of bits would require splitting bytes between channels and/or
pixels. Given the way most programming languages/systems and file
systems work, that requires custom programming. Back in the dark ages, I
encoded and decoded 8 different yes/no bits of data into single bytes
for storage. With the low and dropping cost of storage, I don't see
anybody going to that kind of trouble today.
There is another reason to use 16 bits to work on 12-14 bit source data
relating to the limited number of digital steps available to the lower
values from a linearly converted analog image, but I don't have the time
or brain power just now to try to explicate it. I seem to recall that
there's a good explanation on Luminous-Landscape.
Bottom line - 8 bit throws away data you may or may not want later. If
you don't care, no problem. If you plan to retain the original ORFs and
software to convert them again later, if needed, that will work short
term. Long term, storage in a universal format like TIFF is a better
solution.
Moose
Moose
==============================================
List usage info: http://www.zuikoholic.com
List nannies: olympusadmin@xxxxxxxxxx
==============================================
|