[filmscanners] RE: over resolving scans

["" <snsok@swbell.net>]

Art,

Can you expound on this a bit more in more concrete terms. Let's use the
specific examples of a 4000 dpi scanner and a fine grained film like Velvia.

A 4000 dpi scan would mean that each pixel is thus 6 microns long. What is
the size of the grain in film such as Velvia? Isn't the grain much smaller
than that?

Stan

-----Original Message-----
From: filmscanners_owner@halftone.co.uk
[mailto:filmscanners_owner@halftone.co.uk]On Behalf Of Arthur Entlich
Sent: Friday, October 11, 2002 4:16 AM
To: snsok@swbell.net
Subject: [filmscanners] Re: over resolving scans


In part, you answered our own question.

Scanning resolution isn't just about the necessary input size of a file
to produce a print.

I think a lot of people do not fully understand some of the dynamics
behind capturing a set of random sized and placed dots (grain, or dye
clouds) within a non-random set of squares, and therefore the whole
discussion on grain aliasing gets lost.

Keep in mind that any one pixel can only be one color and luminosity,
and is always square (or rectangular).  Now, what happens when you have
to translate a irregular piece of grain, or overlapped dye clouds in the
case of color film, into that square regular format?

Let's take an example of a black and white film grain the size of the
pixel, but round.  Assuming it was perfectly centered in that pixel (or
CCD sensor) it will be translated into a square element of a certain
luminosity, slightly lower that the actual grain, since some of the
lighter area surround that circular grain will influence the final
luminosity of the pixel.  Already, that grain is enlarged, because the
"round" grain is now filled out into a square, and the luminosity is
also no longer correct.

Now, let's expand that grain to 2 times the size of the pixel element.

Draw a grid 3 units by 3 units, or nine squares, like a tic-tac-toe
board with the perimeter lined as well.  Now, draw a circle centered in
the nine squares, which has a two square diameter.  You should now have
nine squares with a circle pretty much bisecting the outer eight
squares.  Now, imagine that each square can only be one luminosity,
representing the amount of the square which is included in the circle.

The middle square, (Whoopie Goldberg or Bert Renolds, depending on when
you last tuned in Hollywood Squares ;-)) is fully covered by the circle,
so it is 100% gray, or black.  The middle top and side squares would be
about 50% gray since they are covered about that much by the circle, and
the four corner squares would be about 25% gray to represent the amount
of the circle that within those squares.

Two things have occurred.  One, the circular grain is now enlarged to a
3 x 3 square, and secondly, it is represented by 3 different gray
levels, 100% (black), 50% gray, and 25% gray.

As the number of squares required to represent a "grain" increase, the
less "errors" occur and the closer the digital representation becomes to
actually representing the shape and size of the grain, and also the
greater number of squares that are of the correct luminosity.

The way more squares are required to represent a grain is a function of
the size of the grain and the size of each pixel (or the scanning
resolution).  Therefore, the finer the grain is, the finer the scanner
resolution required to represent that grain, but even larger grained
films will show less errors of size and luminosity with a higher
resolution scanner.

Now, imagine what happens when you have irregularly shaped grain, and
overlapping grain or colored dye clouds that make up color film.

You want very many squares to make up one small clump of dye clouds, so
that although each square or pixel can only be one color and luminosity,
it will nearly represent what the film is showing at that location.

The reason the drum scanner provided cleaner, less "grainy" scans that
could handle more USM, is simply because they used a lot more "squares"
(I'm not sure drum scanners use square "pixels", either), but because
less grain aliasing occurred, leading to much finer, more accurate
digital information, meaning much less artifacts when USB was applied.

I have been trying to tell people for a long time that there are more
advantages to be found in a 4000 dpi scanner than that they can make
larger prints, and the reason is because the more component pixels that
make up any one element of the film image, the more accurate in shape
and luminosity the digital representation will be relative to that
original image, and the less gritty the scan will become when USM is
used to restore the sharpness.

And this has nothing to do with no stinkin' dynamic range ;-)

Art



Preston Earle wrote:

> <focus@adnc.com> wrote: "Is 2700dpi "good enough" in many, many cases?
> Sure. Do you lose anything from scanning at higher resolution? Nothing
> except time and hard drive space. (Yes I've read Dan Margulis contrary
> views on the subject - his arguement includes the idea that less
> micro-detail sometimes looks better - obviously, if you agree with that
> there IS a good case for lower resolution, although you can still do it
> yourself better in Photoshop."
> --------------
>
> Isn't scanning resolution more a question of image and print size than
> it is of image quality? As with film and negatives, in general the
> larger the original, the larger a good print can be made. (Boy, that
> sounds awkward--is its meaning clear?) If all I need are 4x6 prints, I
> don't really need 2¼ film over 35mm. Similarly, if I  want 16x20 prints,
> 4x5 film is probably better than 2¼.
>
> For digital images, I only need 1.5 to 2 pixels per output dot/spot for
> most printing applications. A 2800dpi 35mm scan will produce up to
> 12"x18" prints under this rule-of-thumb.  Sure, with a 4000dpi scan I
> could go to 18"x27", but if I'm not going to print larger than 12"x18",
> why scan for higher resolution? If I need to crop an image, then the
> higher resolution may be useful, but just to get more pixel values that
> will to be averaged when the file is printed is wasteful of time, disk
> space, and perhaps quality.
>
> In the good old days of drum scanning, resolutions of up to 10,000 "dpi"
> (or more precisely, scan-lines-per-inch) were available, but images were
> scanned at a resolution to give 300ppi at the largest print-size
> expected for the image. If you guessed wrong and needed the image
> larger, you rescanned it. Every image wasn't scanned at 10,000 "dpi"
> just because you could.
>
> That being said, I've often wondered why drum scanners gave so much
> better "unsharp masking" than Photoshop does. In the late 80's and early
> 90's my printing company had a DS608 scanner that produced absolutely
> gorgeous, very "tight" dark/light halos around images that Photoshop has
> never been able to match. Not being a scanner operator, I don't know why
> this might be the case, but I have wondered if the secret involved
> processing those very fine scan lines prior to converting to half-tone
> dots.
>
> Preston Earle
> PEarle@triad.rr.com
>
> And I still don't know nothing about no stinking dynamic range.
>


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

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