Digital BW, The Print

--- In DigitalBlackandWhiteThePrint@yahoogroups.com, "Paul D. DeRocco"
<pderocco@i...> wrote:
> > From: tvalleau [mailto:tracy@d...]
> >
> > Please don't mention that to the millions of photoshop users who do
> > anti-aliasing every day... ;-)
> 
> Whenever a digital image is reduced in resolution, the image must be
blurred
> before resampling,.......

I thought we were discussing the aquisition of RAW data - light
hitting the CCD, which is an analog function, and does not become
digital until the electrical current hits the A/D converters at the
edge of the sensor.

One has the option of either placing an optical filter in front of the
CCD, which changes the frequences recorded, or grabing the data
without the optical sensor, and interpolating it algorithmically.


> This is the purpose of Photoshop's "bicubic" interpolation
algorithm. If,
> instead, you use the "nearest neighbor" algorithm, which does no
> pre-blurring of the image, you risk aliasing artifacts, if there are
sharp
> edges or fine repetitive patterns in the original image. Once you've
reduced
> the resolution using "nearest neighbor", and you see the resulting
moire and
> jaggies, it's too late to fix them. They _must_ be fixed by
filtering the
> image _before_ resampling, which, again, is the purpose of "bicubic"
> interpolation.
> 
> Similarly, if you start with an optical image, and you want to
sample it an
> turn it into a digital image, you _must_ filter the image _before_
sampling
> it. 

I guess I'm still confused. My camera does not have an optical
anti-aliasing filter. Neither do cameras which have removable optical
filters (such as the Kodak) when the filter is removed... and without
it, they take perfectly sharp photos, which are easily manipulated.

So saying that the ONLY way a digital CCD sensor works is with a
physical optical filter that "_must_" be in front of the CCD, isn't
the case in my universe (or Kodak's apparently.)

> Anti-aliasing isn't particularly sophisticated. 

Well, then I'll admit my stupidity (or age), as I find some of the
mathematical processes to be well over my head, and I'm reasonably
bright otherwise...

see <http://www-ise.stanford.edu/~tingchen/intro.htm> and
<http://www.cs.brown.edu/courses/cs123/lectures/ImageProcess3.pdf>
and the works of Ravjeev Ramanath in the Journal of Electronic
Imaging, including "Adaptive Demosaicking" and "Demosaicking methods
for the Bayer Color Array".

But this is (as can be seen by my failure to understand these simple
formula) getting a bit over my head. It strikes me that we're talking
at cross purposes, perhaps in different languages.

I initially merely want to clarify that which I felt would leave an
incorrect impression in a previous post (the author of which I've
forgotten), which implied that one got 1.5 megs of Red and Blue
information and 3 megs of green, -to Work with- in a digital photo
and that deconstructing a 6 meg photo would somehow reduce to the sum
of those parts... such as 2 1.5 meg channels and one 3 meg channel)
which is a false impression.

The post also introduced "sharpness information" and there is no such
thing coming off a CCD.

Using internal algorithms (which are, as I noted, beyond me) the RAW
CCD information from a 6 meg, 3 color sensor, is 6 megs of red, 6 megs
of blue, and 6 megs of green.

Shooting color for printing B&W benefits from this since differing
frequencies of the light spectrum, and in differing ranged-amounts, is
available as gray information in these three channels, providing the
B&W photographer a great jumping-off point for creating a fully-toned
B&W image.

I guess it's time for me to back out of this conversation, and leave
it by saying that I find shooting a color camera for B&W work, when
one can get a RAW file from it, to be a delight, and capable of at
least meeting, if not exceeding, what I could do in a darkroom 40
years ago...

I've enjoyed the reparte!

Cordially

Tracy Valleau
http://TracyValleau.com