I find this conversation fascinating and hope you will stay the course with all
my (possibly naiive responses.) I'm no expert at all but I have ideas that I
thought were true and I'd like to work them through with you whatever the
On Sat, Jun 08, 2002 at 01:18:17PM -0400, Austin Franklin wrote:
> > Hi Tony,
> > > Number of bits have two main roles. They do indeed represent the
> > > theoretical
> > > maximum density that a scanner could have if the electrical
> > > components were
> > > up to scratch and could use those bits to their full.
> > Not true. You can represent ANY density by any number of bits. I can
> > represent an entire density range with two bits:
> > 00 - at or below dMin
> > 10 or 01 - between dMin and dMax
> > 10 - at or below dMax
> BTW, that should be "at or ABOVE dMax"...sorry.
> > This is a completely valid encoding of density range.
Sure it is. Is that what happens in a scanner though? I thought that the
CCD behaved in a linear manner? Does that not dictate the encoding method
> > I have to run now, so I'll comment on the rest of your post later.
> > Regards,
> > Austin
> And...to add further to this, as I have a few moments... Density values are
> absolute values, just like one foot is an absolute value. They have meaning
> in and of them selves. Someone decided what the exact length of one foot is
> (within a tolerance of course), as well as density values. They mean the
> same thing everywhere...a density value of 1.6 means exactly that in La
> Jolla CA, as well as in Pascagoula, MS. Same with one foot. But, the tonal
> values out of a scanner are not the same everywhere, nor are they the same
> even between scans! The value 136 in La Jolla, CA...does not have the same
> tonality as the value 136 in Pascagoula, MS.
Yep I agree with you here. That density values are absolute values. But
relative density is also commonly used in film sensitometry is it not?
> Scanners are not calibrated TO anything, except themselves. That is why the
> data values you get from the scanner are not the same "density values".
> There is no direct correlation between them, unless you were to calibrate
> your scanner in the same way densitometers are calibrated (even that is
> insufficient, as the sensors used in film scanners have different
> characteristics than the sensors used in densitometers).
OK I have just bought a densitometer which lets me calibrate so that a 0
reading can be set to film base, then readings taken from there. Cannot the
same be done for a scanner? If I scan a piece of film and the film base
gives me a value of 248 (assume 8-bit here just for argument's sake) cannot
I set that as my point from which all my other density values can be
measured? Of course I am making an assumption that the data is behaving
linearly (see http://www.bobwheeler.com/photo/Documents/ZoneDigital.pdf for an
interesting article on this. I'd certainly like your opinion on whether the
information presented seems feasible.) I may not know the absolute density
reading but I would know the density range of the film I am scanning. If I
were scanning a very dense piece of film (say Velvia) then I would need more
than 8 bits to see well into those shadows would I not?
> Now, WHY would anyone in the first place say that number of bits has
> anything to do with density range...because they made some erroneous
> assumptions. Density range is stated in a ratio of "to 1" (:1), and a
> density of 3.6 is 10**3.6:1 or 3981:1 (and someone, at some time, decided
> what the density value of 1 physically is), which the number 3981 requires
> 12 bits to represent in the binary system, if you are going to represent
> every integer value from 1 to 3981...BUT...that's the rub...a value of 3981
> from the scanner is NOT the same as a density value of 3.6 (3981:1), for the
> reasons explained above. And, no, they are not close enough for government
> work ;-)
Yes I agree with you here also but please see my comments above about
relative density readings. Surely one is capable of determining relative
density with one's own scanner? I've made some film charateristic curves
using this method that seem reasonably good.
Running Debian/GNU 2.2 Linux
Unsubscribe by mail to email@example.com, with 'unsubscribe
or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or