> My main area of concern (like Lynn Allen) is in the related area of
> grain-aliasing problems. I think most 2720 dpi users will have
> encountered
> problematic negatives (OK Lynn - and *slides* as well!) where the aliasing
>
> effect becomes horribly obvious. Whereas blurring techniques and GEM-type
>
> software may help with 'normal' grain, I haven't yet found anything that
> helps much with aliasing. For those blissfully unaware, the sort of
> effect
> I am talking about is seen as VERY large grain-like structures, often with
>
> rainbow colors, and usually but not always in the 'thin' areas of
> negatives. I can post (on the web, not to the list) some really awful
> samples if anyone wants them.
>
> While I have seen a number of discussions about techniques to help, they
> are almost always labour-intensive. But I am looking for a simple
> solution. It may be a pipe-dream, but I figure that as *I* can easily
> recognise and describe the difference between grain-aliasing and real
> image
> information, there should be a way for a programming technique or plug-in
> to do likewise.
>
The examples of GEM that I have seen are nothing short of amazing in how
well they reduce grain with minimal reduction in sharpness. However, they
were fro the Minolta Multi II at 2820 dpi (55 lpm) and should have a hard
time differentiating between grain and image detail. I expect the new
Nikons with 4000 dpi (79 lpm) to do much better.
The main thing that distinguishes grain from image is the multi-colored
nature of the grain. Each grain in each color layer is independent, random,
and has high spatial frequencies. The image tends to not be random
(hopefully!), have lower spatial frequencies and little or no content at the
highest frequencies. As I understand GEM, they are using ROC to identify
the individual grains in the digitized image, which are then removed from
the image.
I looked at Minolta's web site for specs on the Multi II (
Section=2 ) and noticed something interesting - the when removing grain the
processing time increases if ROC is turned off. I suspect that they are
performing ROC on the image, removing the grain, and then undoing ROC to
produce the final image.
