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]

[filmscanners] Re: Scanning with too much resolution? (was: PS sharpening...)

Julian writes:

> But I wonder if the loss of detail can be
> remedied by the multistage sharpening/downsampling
> scheme that some people have been advocating.

I think it can; in fact, I think that's the utility of multistep

An unsharp mask actually degrades image quality and introduces artifacts
into an image.  However, the artifacts it introduces create the _optical
illusion_ of sharper detail, thanks to quirks of human perception.  When you
downsample straight to the target size (or when you scan at very high
resolution, in some cases), you get a more accurate image, but one that
contains fewer of the artifacts that call human attention to the existing
detail.  So, depending on how a scanner works, a low-resolution scan might
create the _illusion_ of greater sharpness.  It will never be truly sharper,
but depending on how the samples are picked at the lower resolution, it
might look that way, especially in comparison with a downsampled version of
a higher-resolution scan.

The degradation of sharpening is the reason why I never sharpen images for
archving.  I only sharpen them when preparing a specific image for a
specific final use, and only if that use is for human viewing (which it
usually is, of course).

By doing multiple downsamples and unsharps, the image is significantly
degraded, but the artifacts created serve to create the illusion of greater
detail in a small image that normally couldn't contain such detail.  For
example, a straight downsample of a scene might completely erase small but
potentially interesting details, because they are too small for the new, low
resolution.  That makes perfect sense mathematically, but the details lost
might be important to the composition.  So if you downsample in steps and
unsharp, you exaggerate details in a way that tends to persist as you
approach the target size.  The final image contains considerable distortion
of detail, producing clues that simulate detail in the final image even for
details that should theoretically be invisible.  The net result, for human
perception, is the impression of a sharper final image; and the distortions
that replace real detail may be close enough to reality so that the viewer's
brain replaces them with the _perception_ of real detail.  That is, a
doorknob might not be visible in the final image, but downsampling with
unsharps in steps produces a pixel or two that are different enough in the
final image, and of the right color and placement, to suggest a doorknob to
the viewer's brain--and the viewer's brain calls that odd pixel a "doorknob"
because that's what it expects, even thought it's just an artifact of the
unsharp process.

Photographs are usually intended for human viewing, like other forms of
visual display; and like other forms of display, they can profit from the
idiosyncrasy of human vision to produce more realistic images than should
theoretically be possible given the technical limitations of the medium.

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


Copyright © Lexa Software, 1996-2009.