> This makes no sense to me. Any light that is filters out won't excite
> any silver grains. Otherwise, the filter would have no effect on the
> image at all. You are correct that a filter doesn't increase "noise" on
> B&W film, but I don't think your reason is right. It's simply the grain.
We're making an assumption here of B&W film and not color film which
will yield similar results to digital sensors. When you place a red
filter on a lens it will pass the longer wavelengths but will
eliminate the shorter ones. You will get some light loss in the passed
wavelengths, but the lightloss is calculated as to the loss of white
light.
In the specific case of a Red #25 filter, the level of light reaching
the green and blue detectors is greatly reduced. The specific amount
depends on two factors: The efficiency of the filter and the spectral
response of the green and blue detectors. In the case of B&W film,
each particle of silver grain responds to the full visual spectrum. If
the scene you are photographing is entirely red, there is effectively
no light loss if you place a red filter on the lens. If the scene you
are photographing is entirely green, the red filter will result in an
almost completely black picture.
I was stewing about the contrast issue and came up with the reason for
it. A digital sensor is made of of red, green and blue sensitive
detectors. Depending on the camera and converter algorithm used, each
pixel in the working image is made up of the combined values of a red,
green and blue detector or in some cases two green detectors with one
red and one blue. Assuming for this discussion one red, blue and green
detector in a triad being used for brightness calculation in the
derived pixel, the maximum possible contrast in a true white and black
scene is 100%. But if you place a red #25 filter on the lens, the blue
and green detectors will record a near zero value no matter how bright
the scene is. Therefore, all brightness values are recorded by only
one of the three detectors in the triad. Therefore, it is impossible
for the final converted image to have a contrast greater than 33%.
33% sounds pretty bad, doesn't it? Well, maybe not. The sensor itself
was able to record the full dynamic range, but only on one out of
every four pixels on the sensor or one out of every three pixels in
the triad. But when you RAW convert the file, which derives a pixel
value out of the RGB triad you will need to use a 16 bitdepth to give
yourself the ability to stretch the histogram in post without
posterization. Theoretically, the resulting image shouldn't be any
different than if you shot the scene on B&W film with the red filter
and adjusted the contrast of the scan or the paper-grade to maximize
the histogram. The only thing you should lose in this excercise is
resolution because the red detectors are only 1/4 of the sensor and
also the red detector usually has less effective dynamic range than
the green detectors.
Illustrating the above statement let's keep in mind that Luminance of
a derived pixel is calculated as (R+G+B)/3
Zone-0 recorded with an unfiltered 8-bit digital sensor: (0+0+0)/3=0
Zone-10 recorded with an unfiltered 8-bit digital sensor: (255+255+255)/3=255
Zone-0 recorded with a filtered 8-bit digital sensor: (0+0+0)/3=0
Zone-10 recorded with a filtered 8-bit digital sensor: (255+0+0)/3=85
The method that Moose brings up is the typical method for digital B&W
photography. You shoot the scene without filters and convert the image
to produce a scene that looks exactly how you want the "B&W film" to
see the scene. As B&W films are essentially "daylight balanced" you
may not want to white-balance the scene before applying color filters
in post.
The RAW converted file has 100% of the pixels derived from all of the
detectors of all colors. Therefore, there is no loss of resolution and
you start out with the full dynamic range of the camera file. When you
apply red filteration to the image you are not isolating the red
detectors from the sensor, what you are doing is just filtering the
derived pixels. If a camera, such as the 5Dmk2, is producing unusually
strong noise in the red channel of a DERIVED image, then the problem
isn't in the camera, but in the raw converter. However, if the camera
in question induces a lot of chroma noise in the converted files, then
all bets are off.
AG
--
_________________________________________________________________
Options: http://lists.thomasclausen.net/mailman/listinfo/olympus
Archives: http://lists.thomasclausen.net/mailman/private/olympus/
Themed Olympus Photo Exhibition: http://www.tope.nl/
|