Digital BW, The Print

Truman Prevatt said the following on 4/11/2004 6:11 PM:

> Black and white film is an power sensor. It simply measurses the power 
> (it counts photons) at a location on the film. A CCD is a power sensor. 
> It only counts the photons at a location. The only reason it can 
> measures red or green is there is a filter in front of it that filters 
> one all but one single frequency photons. Light is not only made up of 
> three frequencies - it is a continuum with the visible spectrum being 
> from a little above IR to a little below UV.
> 
> So the question is how well can one predict the actual energy on a pixel 
> per pixel basis using only pixels from there discrete frequencies.
> 
> It is a very easy experiment to run and it has been run many times. Take 
> a CDD without any color filter and measure an image. Measure it with 
> three sensors one with red, green and blue filters. Now take the entire 
> scene and minimize the error of the between a weighted combination of 
> the R, G and B sensors and the output of the CCD without any color 
> filter. There is a set of weights that minimize the error, but the error 
> will not be zero. That is the information lost.

...
Consider another experiment. The image being recorded consists of two 
halves -- one red, one green. The intensity of each half just happens to 
be such the the power sensed by the B&W CCD or film being tested is the 
same for each half. The resulting image will be an even shade of grey. 
Information has been lost, because the sensor throws away all 
information regarding the frequency of the photons. The RGB filter also 
loses some of the frequency related information -- but not all of it.

The information that the eye/brain needs to process an image includes 
both power and frequency. The RGB filter loses some information related 
to the total power over a given area -- but in return it gains some 
information related to the frequency (color). A pure B&W capture loses 
almost all information related to frequency -- the more so if the sensor 
has flat response across the spectrum.
-- 
Clive
http://clive.moss.net