Digital BW, The Print

Mark writes:

> Unless I am missing something, my cameras are rated in RGB pixels
> just as my scanners are and after mixing down to gray scale I have
> the same amount of pixels as I started with...

You are missing something.  Scanners provide all three primary colors for
every pixel.  Digital cameras provide only one color per pixel; the other
two are interpolated from neighboring pixels.  Only 50% of the total
information for green is recorded by a digicam.  Only 25% of the information
for red or blue is recorded.  Overall, the information recorded by a digicam
is only 1/3 the information recorded by a scanner, for a given number of
pixels.

When you convert to grayscale, this results in a loss of quality.

Additionally, RGB conversions to grayscale always provide less image quality
than direct recording of the original scene in black-and-white, whether the
original method of image capture is digital or film.

> The real way to look at this is that the sensors are
> throwing away a lot of b&w resolution to interpolate
> to RGB in the camera ...

The loss of resolution is not in the interpolation; the resolution is
already gone by that point.  The loss is in the initial filtering to red,
green, and blue.

> ... and *if* someone were to make a dedicated b&w digital
> camera using current sensors it would have much
> higher resolution than current color models.

This has been done (by Kodak), but I don't believe that model is still
available.  Digital sensors provide excellent results in black and white, if
they are dedicated to that purpose (that is, no matrix color filter over the
sensor).

> The biggest draw back to a dedicated b&w camera is that,
> just like film, it would have only one response to a full
> spectrum of visible light and filters would have to
> be used etc. etc.

Yes.  That's also true of RGB digital cameras vs. color film, too, of
course.

> Mixing RGB down to b&w gives you a lot of
> freedom when you are "processing" your images at the cost of
> resolution...

Not really.  You lose a great deal of information in the initial conversion
to RGB in the camera, and it's much more than just resolution.  Many
black-and-white photographic effects are impossible in conversions from RGB,
because the original information just isn't there.  You cannot simulate many
filters with RGB conversions, for example.

> ... of course, soon you will have more resolution than you
> will ever need and it won't matter if you are shooting a color
> sensor...

Not any time soon.

> Sure, if you shoot Techpan you will get great scans... you will get
> great results even with a 2880ppi film scanner and 35mm... but even
> at 80 asa in TD-3 that is too slow for lots of stuff.

Portra 400BW gives results that are similar (if not equal to) Technical Pan,
and it does it with 16 times greater speed.  In medium-format and night
shots in particular, Portra 400BW is amazing.

Umm, we are mixing pixels with sensor elements here.  Sure, the 
typical CCD does not provide a true rgb pixel for every sensor 
element, it has to interpolate one from adjoining elements which 
means your effective pixel area is larger than any individual sensor 
element, but that doesn't mean that information is being "thrown 
away", just that your effective pixel size is not equal to your 
individual sensor element.  I guess we are just looking at the same 
system and characterizing it differently...  Sure, a true and 
instantaneous scan of each rgb channel would be preferable, but in 
real life there are so many variables in film scanning that the 
benefits are probably lost... 

When you convert to gray scale you don't necessarily "lose" quality 
over your rgb information, you are just getting there in a terribly 
inefficient manner... your effective pixel size being so much larger 
than you sensor element...

comparing scanners and digicams and making a blanket statement that 
you get 3 times the information from a scanner is really misleading 
at best.  They are different tools and they are all working at 
different resolutions and in a different manner... and that doesn't 
even address the A->D issues.

> Portra 400BW gives results that are similar (if not equal to) 
Technical Pan...

Then why the heck would anyone shoot Technical Pan film then????  I 
can't even conceive how anyone can make this statement?  How???  Are 
you hitting everything with a "grain filter" or something before 
looking at your results???  How are you developing your TP???? ... or 
do you have some super-secret magic way of developing 400BW?  If you 
can show me how to get TP results from 400BW I will worship you 
forever:)

> The loss of resolution is not in the interpolation...

no, it is in the design... I thought my statement was clear enough 
here... apparently not... sorry.

I don't think there is a real argument here anyway... just 
semantics...

mark

--- In DigitalBlackandWhiteThePrint@yahoogroups.com, "Anthony 
Atkielski" <anthony@a...> wrote:
> Mark writes:
> 
> > Unless I am missing something, my cameras are rated in RGB pixels
> > just as my scanners are and after mixing down to gray scale I have
> > the same amount of pixels as I started with...
> 
> You are missing something.  Scanners provide all three primary 
colors for
> every pixel.  Digital cameras provide only one color per pixel; the 
other
> two are interpolated from neighboring pixels.  Only 50% of the total
> information for green is recorded by a digicam.  Only 25% of the 
information
> for red or blue is recorded.  Overall, the information recorded by 
a digicam
> is only 1/3 the information recorded by a scanner, for a given 
number of
> pixels.
> 
> When you convert to grayscale, this results in a loss of quality.
> 
> Additionally, RGB conversions to grayscale always provide less 
image quality
> than direct recording of the original scene in black-and-white, 
whether the
> original method of image capture is digital or film.
> 
> > The real way to look at this is that the sensors are
> > throwing away a lot of b&w resolution to interpolate
> > to RGB in the camera ...
> 
> The loss of resolution is not in the interpolation; the resolution 
is
> already gone by that point.  The loss is in the initial filtering 
to red,
> green, and blue.
> 
> > ... and *if* someone were to make a dedicated b&w digital
> > camera using current sensors it would have much
> > higher resolution than current color models.
> 
> This has been done (by Kodak), but I don't believe that model is 
still
> available.  Digital sensors provide excellent results in black and 
white, if
> they are dedicated to that purpose (that is, no matrix color filter 
over the
> sensor).
> 
> > The biggest draw back to a dedicated b&w camera is that,
> > just like film, it would have only one response to a full
> > spectrum of visible light and filters would have to
> > be used etc. etc.
> 
> Yes.  That's also true of RGB digital cameras vs. color film, too, 
of
> course.
> 
> > Mixing RGB down to b&w gives you a lot of
> > freedom when you are "processing" your images at the cost of
> > resolution...
> 
> Not really.  You lose a great deal of information in the initial 
conversion
> to RGB in the camera, and it's much more than just resolution.  Many
> black-and-white photographic effects are impossible in conversions 
from RGB,
> because the original information just isn't there.  You cannot 
simulate many
> filters with RGB conversions, for example.
> 
> > ... of course, soon you will have more resolution than you
> > will ever need and it won't matter if you are shooting a color
> > sensor...
> 
> Not any time soon.
> 
> > Sure, if you shoot Techpan you will get great scans... you will 
get
> > great results even with a 2880ppi film scanner and 35mm... but 
even
> > at 80 asa in TD-3 that is too slow for lots of stuff.
> 
> Portra 400BW gives results that are similar (if not equal to) 
Technical Pan,
> and it does it with 16 times greater speed.  In medium-format and 
night
> shots in particular, Portra 400BW is amazing.