If I understand what you are saying, I think that I cannot agree with your
explanation. Your analogy appears to be confounding halftone dots with
halftone cells. Moreover, it is not necessarily the case that either
translate one-to-one into pixels or into samples. Also I believe that if
your analogy was correct, the 2000 dpi would represent a halftone cell
consisting of two dots while the 4000dpi would be a halftone cell with 4
dots, such that there would be 2000 cells per inch (or 4000 halftone dots
per inch) versus 4000 cells per inch (or 16000 halftone cells per inch).
First, technically there is a difference between dpi which is usually used
in reference to resolution in terms of halftone dots or cells per line per
inch ( or lpi -lines per inch in printing press terms) on a printed page
versus ppi which is used with regards to pixels which typically refer to
resolutions in terms of picture elements in a monitor display versus spi or
samples per inch which refers to resolutions in terms of the number of
samples captured by a capture device such as a scanner or digital camera
from the original subject. Often and usually wrongly, these measures and
terms are used interchangabley as if they were identical or equivalent to
one another.
Second, the key factor in determining the quality and equivalance of scanner
resolutions is the difference between the native optical resolution of the
scanner (whatever terminology is used to define the units per inch) and
interpolated resolutions or software generated resolutions of the scanner
(whatever terminology is used to define the units per inch). The former
comprises the actual scanner resolution as opposed to some mathematically
generated derivative of the actual resolution.
Third, with respect to output resolutions and the original question, the
quality of a scanner and its output is as much determined by the bit depth
of the scanner ( i.e., the dynamic range of tone that the scanner can
capture and recognize and the capacity of the scanner to recognize tonal
distinctions within that dynamic range and discern or differentiate those
distinctions from noise) and the quality of the scanners design, sensors,
and hardware componets as by the optical resolution if it capable of
capturing and out putting at. Thus, the stated resolution differences
between scanners may be irrelvant in terms of the quality of the output.
Furthermore, the amount of resolution needed to do a quality capture and
output will depend on the size and type of original that is being scanned as
well as the pruposes to which the scan is being used. For example, one does
not need as much resolution for reflective originals as for transmissive
originals or for large originals as opposed to small originals. To use a
scanner to scan an 8x10 sheet of film at 4000spi is overkill unless, for
instance, one is going to enlarge the captured image to billboard size or
crop out and enlarge into a 16x20 inch image only the head of one person in
a large group shot, holding all other scanner spec equal.
Fourth, the size of the output resolution, holding other variable constant,
becomes important when one is scanning small 35mm originals which will be
enlarged 5 or more times full frame upon printing or which will be cropped
and enlarged to a size of 5 or more times the size of the 35mm original full
frame. Here, you want an optical resolution high enough so that you can
enlarge the image size which will effectively reduce the resolution and wind
up with an optical resolution of 200 - 300 units (whatever terminology is
used to define the units per inch) without resorting to interpolation.
Thus, 4000 spi scanner will allow one to enlarge the scanned frame larger
without resorting to interpolation (e.g., resampling upward) than would be
the case for a 2000 spi, assuming both scanners resolutions are optical
resolutions. This would be true for originals whatever their size; but more
than 2000 spi may not be needed to scan for instance a medium format frame
and 1200spi may not be needed to scan a 4x5 or larger film frame. This is
also the case for reflective originals scanned on flatbeds with respect to
optimum resolutions in relation to final image sizes; but with respect to
quality of scans, anything more than 600-1200 spi for outputs 1-3 times the
size of the original is usually unnecessary since the dynamic range of most
reflected originals is much narrower than for transmissive originals and
details tend to be more blocked up on the extreme ends of the histogram for
reflective originals than for transmissive originals.
filmscanners_owner@halftone.co.uk wrote:
> At 01:37 PM 3/25/2004 -0600, you wrote:
>> I'm a bit perplexed at what the dpi means on a film scanner. Trying
>> to compare apples to apples, will a 4000 dpi Brand X film scanner in
>> theory produce a better quality image outputted than a 2000 dpi
>> Brand X scanner, given that the output resolution is the same, say
>> 1600 x 2400 pixels?
>>
>> Or does it simply mean the 4000 dpi scanner will output a much larger
>> image than the 2000 dpi model?
>>
>> Thanks for clearing this up,
>> Bill
>
> Bill, try to think of it this way. If a 2000-dpi scanner renders a
> given negative as an image composed of a single dot (".") then a
> 4000-dpi scanner would render the same negative as an image composed
> of four dots, two over two (difficult to display on an ASCII list,
> but, if the formatting cooperates, something like this):
>
> ..
> ..
>
> So, if you blew up both images in Photoshop, say, to where you were
> able to work with individual dots/pixels on the screen, the image
> from the 2000-dpi scanner would afford you just one dot (screen
> pixel) of scanned information to edit (manipulate) from any given
> area of the scanned negative whereas the image from the 4000-dpi
> scanner would afford you four dots (pixels) of information to edit
> (two dots across and two dots down).
>
> This difference (fourfold) often represents a significant difference
> for "critical" editing work with regard to printing pictures (the
> kind we actually hang on our walls) at larger sizes (I'm not talking
> 4x6 stuff here). For purposes of display of the final image online
> somewhere, this difference, when boiled down to real-world
> constraints, may or may not be relevant (visible). It just depends.
>
> An example of where this added (4x) dot/pixel information comes in
> handy: let's say you've sharpened (USM has been a recent topic of
> interest) an image with someone's face the center of attention, and
> you're left with some unattractive artifacts around the nose. It
> would be easier to go into (blow up) the 4000-dpi image and "smooth
> out" (or if you will, more attractively _blend_) those nose artifacts
> than it would be to do the same thing with the 2000-dpi image for the
> reason that with the 4000-dpi image you would have four times the
> information (four dots/pixels) to work with as opposed to just one.
> It's impossible "blend" one dot/pixel. All one can do is change any
> given unit to something else, yes? On the other hand, with _four_
> dots/pixels to work with the possibilities for change are relatively
> much more various.
>
> Hope that helps.
>
> And yes, I often do go into my work and manipulate intransigent
> dots/pixels one by torturous one. (I may not be especially talented
> but I always get an A for effort. <g>)
>
> Tris
>
> --------------------------------------------------------------------------
--------------
> Unsubscribe by mail to listserver@halftone.co.uk, with 'unsubscribe
> filmscanners' or 'unsubscribe filmscanners_digest' (as appropriate)
> in the message title or body
>
>
>
> ---
> Incoming mail is certified Virus Free.
> Checked by AVG anti-virus system ().
> Version: 6.0.617 / Virus Database: 396 - Release Date: 3/9/04
---
Outgoing mail is certified Virus Free.
Checked by AVG anti-virus system ().
Version: 6.0.617 / Virus Database: 396 - Release Date: 3/9/04
----------------------------------------------------------------------------------------
Unsubscribe by mail to listserver@halftone.co.uk, with 'unsubscribe
filmscanners'
or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or
body
