[filmscanners] RE: 8bits vs. 16bits/channel:cantheeyeseethedifference

["Austin Franklin" <austin@darkroom.com>]

Hi Frank,

> That is really cool if PS does this, but Franklin flatly denies it.

I believe I understand what Roy is saying, and I certainly can believe that
PS does something algorithmically, like round up or down (or a host of other
algorithms), instead of just lopping off the 8 LSBs, that I have no problem
with.  It's whether what it is doing is in fact dithering.  Simply rounding
up or down is not dithering, or using a deterministic algorithm (one that
gives the same output given the same input).  It would have to be inducing
random noise into the data for it to be dithering, that's what dithering is.
I believe is possible, but improbable that PS dithers the LSBs as Roy
contends.

> He
> says PS does a simple truncation, which would be pretty stupid if it
> does, and I don't think PS is known for the stupidity of its algorithms.

That actually is speculation on my part, believing that any other method
does not improve an image significantly, though they very well may use some
other deterministic method.  Again, what I am questioning is if it is
dithering or not, not trying to specify the algorithm it's using for doing
the conversion.

Simple truncation is really not that bad, another method would be simple
rounding...and you are really not going to get many values that fall EXACTLY
half way between two values (as was used as an example).  What rounding
gains you is a 1/2 shift (what was greater than .5 and less than the next
whole value now becomes the next whole value).  I don't believe image wise,
that is a significant improvement over simple truncation.

> However, I suppose Franklin is right in saying this dithering loses
> tonal resolution, not spatial resolution. Sigh...I suppose he can't
> always be wrong...

Yeah, ha ha...Frank.  If I'm wrong I admit it, and always have, but since
that doesn't happen much, it doesn't happen much  ;-)

Let's see what Roy comes up with to prove if it's dithering or not.  He said
he can demonstrate it, and I'd like to see if he can show that.  I believe
what is going to happen is he's going to show it's algorithmic and
deterministic, not that it's dithering.

Whether it's dithering or now is easy to determine.  Take a 16 bit image,
and change it to 8 bits, save it, and do it again and save it to a different
file. Compare the image data from the two files, whether an FC would work or
not, I don't know, as I don't know if there's a timestamp or something like
that in the data file.  If the image data is different, then it's dithering,
if not, it isn't.  But, you have to make sure you did the exact same thing,
and that the file only contains image data, in order for this to be a valid
test.

Regards,

Austin

----------------------------------------------------------------------------------------
Unsubscribe by mail to listserver@halftone.co.uk, with 'unsubscribe 
filmscanners'
or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or 
body