Sunday 20 August 2006

Scanning 35mm film


About why to design before implementing

Yesterday I gave a shot at trying to scan some of the 35mm negatives of my old Black&White photos using a 1200 DPI flat-bed scanner. It failed misserably:
  • The scan showed severe Moiré and other kinds of noise in the pictures.
  • The sizes of the TIFF files were enormous.
  • At 1200 DPI the flat-bed scanner picked up dust and stains from the glass.
Although it was just an experiment, the terrible results led me to do some research. Unless otherwise stated, the following numbers come from the article about Digital Photography at Wikipedia.

Resoluiton of 35mm film

The Wikipedia article says that a DSLR resoluiton of 6 to 14 megapixels aproximates that of 35mm film, and that one of 16 megapixels exceeds it, so lets asume that the resolution of 35mm film is 14 megapixels.
To compare with the scanner resolution we need to translate that to DPI. Megapixels are calculated as the product of the vertical and horizontal resolutions:
M=x*y
The Aspect Ratio of 35mm and DSLRs is known to be (3:2), or 1.5. Thus:
x=1.5*y
y=x/1.5
and
M=x*(x/1.5)
M=x^2/1.5
x=sqrt(1.5*M)
So, at 14 (actually 14.7) megapixels, the number of horizontal pixels in 35mm film is about 4693. If we divide that by 1.38" (35mm translated to inches), we obtain 3405 DPI for 35mm film.

Required resolution

With that result it is not surprising why scanning 35mm with a 1200 DPI scanner produces terrible results: 65% of the information in the negative is not captured. Even more, according to Information Theory, the sampling density must be the double of the information density, which means that the scanner should have a resolution of at least 6810 DPI to sample a 35mm negative without bias.
Guess what? The resoluiton of specialized 35mm film scanners is of above 7000 DPI…

No comments: