[filmscanners] Re: Correct/best methods of scanning

["" <anthony@ohalloran.co.uk>]

> As an example, one of them called
> for scanning a pure black, thick line drawn on a pure white background, and
> then resample/downsize via bicubic or whatever: the borders turn grey,
> which is a color not present in the original image. He concluded that
> downsizing blends colours and creates artifacts (but, since he's operating
> an Imacon Flextight maybe he doesn't really need to increase detail).

There are quite a few things I would like to say here, these can be
quite hard to explain in a few lines but i'll have a go:

If you scan a perfectly sharp black line the boundary pixel between the
line and background will always be grey. As the sensor moves along it
will eventually move into a position where some of it is over the
background and some over the line. This will produce a grey pixel. The
intensity will depend on how much background/line is under the sensor.
Of course the exception to this is if the boundary on a particular scan
contains less that 1/256 background (assuming 8 bit scanning mode). This
pixel will then be pure black.
Now the above makes some over simplifications about sampling. The sample
is not made using a square mask with every part of that square
contributing equally to the output pixel. Therefore you gonna get grey!
ALWAYS!

Going back to the folks resizing the (let's assume) perfectly sharp line
above. The process is essentially identical to what i described above so
yes the edges will get grey.

Now, fortunately for us we have unsharpmasking filters (USM) available
to us which are get rid of our soft edges. This is what they should be
used for. If you down sample images you always need to USM for the
reasons discussed above.

> Other colleagues are scanning at maximum resolution -to capture all the
> detail the scanner can resolve- and downsizing later. Here, the point is to
> apply some techniques recently discussed on the list via wise sharpening
> filters between resizings.

You get the best final result if you scan at full optical res, down
sample, USM. If you scan at 4000dpi then down sample to 2000dpi you are
taking info from 4 pixels and putting it into one. Far better than just
randomly picking one of the four original pixels. This greatly helps to
keep uniform colours uniform!

Anthony

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