[filmscanners] Re: PS sharpening

["Julian Vrieslander" <julianv@mindspring.com>]

On 8/11/02 9:33 AM, "Anthony Atkielski" <anthony@atkielski.com> wrote:

> Al writes:
>
>> Maybe I have missed it in an earlier post but, if
>> you are using your normal technique of halving the
>> image size, what are the unsharp mask settings you
>> use as a default?
>
> Strength of 98, radius of 0.7, threshold of 2.  Of course, this is a highly
> subjective setting.  I do note that very small images usually require less
> unsharp masking than very large images to get visually similar results, but
> since the distinctions are small, I usually use this one setting for
> everything.  If a small image looks too pixellated after the last downsample
> and unsharp masking, I undo the unsharp masking.

I ran some tests of the multistep downsampling/sharpening process, comparing
it with my usual method.

I started with a 4000ppi scan of a sharp Provia slide.  I opened it in PS7
and downsampled in one step by a factor of 8, using bicubic.  Then I applied
unsharp mask, adjusting settings for subjective best result on my monitor:
amount = 150%, radius = 0.3, threshold = 0.  The image was saved to a TIFF
file.

Then I started with the same original scan and downsampled in three steps,
halving resolution at each step, with bicubic.  After the first two
reductions, I applied unsharp mask, using Anthony's suggested settings.
After the third reduction, the image was the same size as the image which
was reduced in one step.  If I then applied unsharp mask with Anthony's
settings, the image was obviously oversharpened, with halos, etc.  I could
play with the sharpening settings to make the multistep image look similar
to the one step image.  But I could not find any settings that made it look
clearly better.

So I am not yet convinced that this multistep process provides any
advantage.  A year or two ago, I read a report from another person who did
some more elaborate tests than mine, and reached a similar conclusion: that
the multistep downsampling/sharpening did not produce a better result than
a single step method.  But I am still open to being convinced.  Am I doing
something wrong?  Are there specific circumstances where the multistep
method shows its advantage?

Since I am judging the results on a monitor, one could argue that the
display is not adequate to show the differences.  But, for me at least,
downsampling is only necessary when I am preparing images for display on the
web.  My images all come from 35mm film scanned at 4000ppi.  If I am
printing, I use an Epson 1280 inkjet.  I resize in PS with the "Resample
image" option disabled, letting the resolution fall where it may.

--
Julian Vrieslander <julianv@mindspring.com>

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