Digital BW, The Print

CDTobie@... wrote:
> In a message dated 4/30/06 9:32:58 PM, tyler@... writes:
> 
> 
>> A lot of people don't understand this. I'm continously asked what dpi file 
>> is needed from a
>> given film size for a given print size. Their assumption is that since it's 
>> commonly thought
>> 360 dpi at print size is desirable, then the scan must reflect that.
>> For a large print from 35mm, that requires a scan at higher dpi than many 
>> scanners can
>> deliver. But more importantly, there is no image detail resolved to that 
>> degree or anything
>> close to it, given lens and film capability. In fact, you are more likely to 
>> simply over-
>> resolve grain, to a distressing visual degree much greater than would have 
>> apeared at a
>> same size print made through an enlarger.
>> There are so many case by case factors that apply, it's very difficult to 
>> make
>> generalizations abou these discussions.
>>
> 
> Which brings us back to my concept of differing degrees of grain scanning. 
> When resolution first starts showing variation caused by grain, it does not yet 
> resemble grain, so when scanners were of that resolution the concept of 
> "digital grain" in scans arose. Once grain could be articuated more realistically by 
> affordable scanners, users got quite enthused about scanning to the grain. 
> But you can take it even further with higher rez scanners. Such images are all 
> grain, and the images can't effectively be sharpened, any sharpening just 
> accents the grain. Touchup is also difficult, as it must deal with maintaining 
> grain pattern. At some point the idea of shooting digitally, doing all of the proc
> essing and sharpening, and then adding synthetic grain to the image, to cover 
> Tyler's digital fuzziness issue, starts to seem easier. Higher image 
> resolution, be it in film or digital, sounds like a better idea to me!

David,

Strange enough I find touch up on grainy scans not that
problematic. The stamp and 100% opaque works quite well when
the copied area is selected carefully. With wet mounting and
clean working there isn't much retouching needed either.

More problematic is the contrast shifts you get in sharpening
+ the reverse, more pronounced grain with contrast increase.
There should be a tool or action that balances the sharpening
with the contrast reduction and reverse. I understand that it
is very film/grain/scan resolution specific. I mentioned that
problem before, the tone scale is no longer dictated by grain
size + grain distribution but by the pixels that describe the
grain particle. It is a catch 22 problem. You will loose the
grain with higher resolution film and the same scanning
resolution (usually the max of that scanner) but that doesn't
mean the image detail is increased as well, the camera lens is
more likely the limiting factor in that method. That path
followed to the extreme would generates the same void
upsampling Tyler describes.

Drum scanners have hit that problem long before the CCD
scanners were available. The aperture control on drumscanners
is a hardware solution (combined with software) that can deal
partly with this. It distills the last of the grain 
information while eliminating actual grain in the scan. 
Depending on the resolution of the camera lens and what is 
left of that in the grain you could create the last real 
information or get the same voidness described before.
Similarity is found in grain reduction + sharpening software
or in the over-sampling of the Epson scanners. It can work if
the grain distribution and grain size still holds scene detail
at that level but doesn't if that isn't case.

I already indicated that film material for scans could be
improved. I also wonder whether some nifty software algorithms
could deal with grain structure translations to the
pixel matrix. Like sub pixel anti-aliasing and hinting
preserves curves, angled lines, serifs, of fonts on a pixel
structure one could think of shifting grain particles slightly
to fit the pixels better. Crazy idea probably but the same
must have been said when font designers added scaffolds,
skeleton  lines etc to fonts to preserve their quality on bad
displays and that with the help of slow processors 15-20 years
ago. A grain particle isn't a very complicated shape and
computers are good in stacking already.

I wonder at what stage artificial grain gets as annoying as 
upsampling artefacts are. I have seen enough annoying samples 
in ad designs already.


Ernst
-- 

                    --
           Ernst Dinkla


www.pigment-print.com
(         unvollendet         )