[filmscanners] Re: Pixels and Prints

["Arthur Entlich" <artistic-1@shaw.ca>]

I'm with Austin on this one...

Yes, the scanner would be hard pressed to fully accurately represent
every grain (dye cloud) unless if was extremely high resolution.
However, within its resolution, it accurately represents the "average"
hue and luminosity that the film represents in that pixel location.

The Bayer pattern doesn't average, it interpolates.  It "guesses" the
value based upon surrounding pixels, and it can be off a fair amount
during transitions and drastic hue changes.  Each location, each pixel,
in the film scanner represents an averaging of the pixel dimensions of
the film information it represents.  Not so with Bayer patterns.  Any
one pixel only "knows" one color component (R,G or B) and luminosity of
that area, nothing more.  All other information is determined by
assumptions about the neighboring pixels values, which also lack
information about the other component colors.  50% of the pixel captures
in the Bayer Pattern have no "knowledge" of the red and blue component
in their color.  25% have no true input for the red or green and 25%
have no true input for the green and blue component. In other words,
clearly 75% of the pixels do not have a defined red component and the
same for blue.  It's actually amazing that the concept works at all ;-)

Every pixel on a film scanned image has all three color components
accurately expressed within the limitation of its resolution.  It
literally samples every pixel it represents in all three component
colors, averaging them within the area of that sample.

The one advantage of the digital camera is that the capture media is
identical to the sampling in X/Y dimension, so there is no averaging of
information within the sampling area. In its native resolution, one CCD
sensor equals one pixel dimension, and therefore the color and
luminosity information within the area is already "averaged" by the
sensor, from the image projected on it by the lens. where grain (dye
clouds) is random is shape, location and overlap.

Art

PS: I'm not sure why Austin and I are coming up with different
percentages on the red and blue accuracy values.  He indicates 66%, and
I'm coming up with 75%.  I'll defer to him, because this is his area of
expertise, but I don't understand what it is I'm doing wrong in my
thinking.  As I understand it, the Bayer pattern uses a system with 50%
green, 25% red and 25% blue sensitive sensors (using color separation
filters).  Since neither the green pixels (50%) nor can the blue (25%)
can accurately detect the red component, doesn't that mean 75% of the
the locations are interpolated for the red component?


derek_c@cix.co.uk wrote:
> Of course each pixel of a scanned film has all three colours faithfully
> reproduced.
>
> The interesting question though is what that pixels's actual colour was?
>
> Unlike a camera, a film scan records something that has already been
> sampled into RGB, that's what film does!
>
> Yes the film grain is much smaller, and randomly scattered to boot.
>
> Sample aliasing means that each scanned pixel overlaps a number of film
> grain dye images, each an individual R G or B. So the pixel itself is
> effectively just an average like a Bayer filter pixel quad is an average.
>
> So it seems to me that there is no real moral high ground there, film
> scanning is just as much an averaging process as a Bayer camera.
>
> In article <NABBLIJOIFAICKBIEPJJMEMNIHAB.austin@darkroom.com>,
> austin@darkroom.com (Austin Franklin) wrote:
>
>
>>Roger,
>>
>>So what if it's second generation?  Unless you can analyze the fidelity
>>of
>>it to make claims from, that's simply an argument that has no teeth.
>>
>>Fact is, digicam pixels have some %66 of the red, %66 of the blue and
>>%50 of
>>the green data interpolated.  Scanned film data does not.  It has all
>>three
>>color values as original information.  So, second generation or not, the
>>fidelity (which is what is important) of the data from scanned film far
>>outweighs digicam data of the same "resolution".
>>
>>How good the scanned data is, depends a lot on how good the original
>>film
>>image is, as well as how good the scanner/operator is.  Not all scanners
>>scan 4000 PPI the same.
>>
>>Even if you recorded Ozzie live with your 8 track tape recorder, my nth
>>generation CD will have a far higher fidelity.
>>
>>Regards,
>>
>>Austin
>
>
>

----------------------------------------------------------------------------------------
Unsubscribe by mail to listserver@halftone.co.uk, with 'unsubscribe 
filmscanners'
or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or 
body