Apache-Talk @lexa.ru 

Inet-Admins @info.east.ru 

Filmscanners @halftone.co.uk 

Security-alerts @yandex-team.ru 

nginx-ru @sysoev.ru 




      :: Filmscanners
Filmscanners mailing list archive (filmscanners@halftone.co.uk)

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: filmscanners: Sharpening scanned images for printing

On Fri, 07 Dec 2001 23:20:32 -0500  SKID Photography (skid@bway.net) 

>  If you claim that
> 'scans', are not of the grain of the film, I don't understand where the 
> scanner is getting it's information
> from.

For once I agree with Austin :-)

The scanner sees only luminance and colour, not image detail or grain 
boundaries. It is essentially blind to all contrast transitions which have 
a higher frequency than half the CCD pitch.

With film, image detail *always* has a much lower frequency than the grain 
which constitutes the image. Grain acts as an irregular halftoning, and 
although the implicit resolution capability is empirical rather than 
calculable, the Nyquist limit still holds for film. That is why fast film 
can record detail less well than slow films.

It is quite possible to scan and transpose this image detail without 
recording grain boundary information. 

In fact the CCD acts as a low-pass filter, which adds some distortions to 
frequencies close to the filter cut-off point (as aliasing).

If a grain image at the CCD is smaller than a pixel, the pixel averages 
the values it is presented with and an aliased representation is the 

If a grain image is larger than a pixel, this  still happens at pixels 
where there is only a partial overlap - the edge of the grain will produce 
an aliased pixel value. Only fully-overlapped pixels will be free of 
aliasing errors.

Whatever you see in a scan as grain is an aggregated, aliased 
representation of grain mapped to the physical grid of the pixels. Only if 
pixels are very much smaller than grains can this be an approximately 
accurate resolution of the grain. Existing CCD scanners just don't get 
near that. In fact I have seen drumscanned images of 4,000ppi, 8,000ppi 
and 12,000ppi from ISO100 Ektachrome where each successive iteration 
produced marginally more fine image detail yet significantly more 
accurately rendered grain. 

Even if you had perfect optics and could scan at 100,000ppi, you'd still 
get aliased, fuzzy pixels instead of grain boundaries. They'd just be 
very, very small pixels.

The same is true of actual image boundaries, except being of a lower 
spatial frequency, the distortion is a lower percentage. Losses are 
inevitable in getting stuff from the analogue to digital domain.

Scanning is like duping. Dupe onto a film which is coarser than the 
orginal at 1:1 and you will lose image detail and sharpness. Dupe at 10x 
and you will preserve most of the sharpness, but the original grain will 
be compounded by the copy film grain. Use a fine-grain film instead, and 
everything is less worse. However losses are inescapable, and 
acceptability a judgement call.

So does any of this matter? That depends purely on output size and 
technology, and viewing distance, plus how good your eyes are :)


Tony Sleep
http://www.halftone.co.uk - Online portfolio & exhibit; + film scanner 
info & comparisons


Copyright © Lexa Software, 1996-2009.