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: Unsharp mask was Re: filmscanners: Getting started question

I find you comments about analogue "feel" very interesting , as I just
wrote a reply in the "other" scan list that I think I will post here as
a result.  I think this is called convergence. ;-)

Actually, I just realized, that Dave wrote the comments I am replying 
to in both lists...

Humans do not like rigidly "gridded" anything.  There is a random 
element in nature, and in us, and we like it.  Noise is random,
photo grain is random.  Digital sampling is rigid, pixel positions are
rigid.  We are analogue.  Neither is accurate, but we are more
comfortable with analogue because we prefer randomness, and our eyes and
ears are analogue and create all sorts of randomness.  

We require "fuzzy logic", and we incorporate it into our machines
because they can "think" better (as we do) that way.  Since we program
them, we tend to use our type of logic, which in turn probably slows
them down, but by introducing the "fuzzy" part, we make them more able
to function as we do, which makes them, like us, skip over the less
important details.

Humans tend to become more proficient at tasks by learning or training
ourselves to ignore most of the input we receive, to "narrow focus" on
only that which is relevant to complete a task.  Left to their own,
computers analyze every piece of information they receive without being
able to selectively "block out" the unimportant stuff.

Go to a cocktail party, and without moving your position, follow
conversations in different parts of the room.  Our brain allows us to
amplify certain vocal tones, frequencies and spatial placements, while
diminishing others.  Now, try to design a machine which can do the same
without further human intervention, ---call me 50 years from now when
you have it worked out. ;-)  Ask anyone who wears hearing aids how
annoying it is to have all the sounds in the room amplified, and having
lost control over this selective hearing.

One of the reasons inkjet printers seem to translate images so well (to
our liking) is because they use random dithering techniques.  We like
sub-threshold noise, and now that I've made enough of my own (noise),
I'll end this posting ;-)


Dave King wrote:

> Or one can use the simple approach of sharpening "grain" (or whatever
> it is:) with no regard to individual image detail.  I prefer to look
> at the grain in an area of no detail in fact, at 100% at the final
> print size.  I've been using 75% at .8 radius, 0 threshold for most
> things with the Agfa T-2500, and sharpening the original scan once and
> then again if the image has to be interpolated up considerably for
> large print sizes.  This seems to me to be closest to an "analogue
> feel" in the final print, but it does throw away one of the advantages
> of digital that many people love.
> I've found in comparisons of Genuine Fracticals to Photoshop bi-cubic
> (in two stages when much upsampling is needed) that Genuine Fracticals
> introduces edge sharpening effects that are not "analogue like", and I
> prefer bi-cubic.  Also, in areas of complex detail (grass for
> example), GF can get confused and make a bit of a mess of things.
> Even so, I've found the differences between GF and bi-cubic to be
> pretty subtle in the final print.
> Dave


Copyright © Lexa Software, 1996-2009.