Digital BW, The Print

>Peter said: You might find it worth downloading the trial version 
>of Photozoom Pro in that case as it seems almost as good. But if 
>you already have QImage its probably not worth bothering.

I just downloaded and tried the Photozoom demo. Results are below in
the list.  Here's what I'm doing and what I've tried so far.

The Challenge: upsize some 1.2 mp JPGs that were shot with a Casio
Z-50, a pocket sized 5mp digicam.  Upsize to 8mp.

What I've tried:

1) PS-CS/bicubic - surprisingly good results, way beyond my
expectations.  Everything else is compared to this.

2) PS-CS/bicubic Smooth - not bad but compared to 1) is too soft to
recover w/ more USM.

3) PS-CS/bicubic Sharper - increased contrast too much & compressed
shadows, plus halos.

4) Qimage/Pyramid Print-To-file, at various sharpen settings - lots of
patterned atrifacts and halos.  Not even close to 1). 

5) Qimage/most other algorithms - various degrees of artifacts and
halos.  Nothing anywhere near 1) above.

6) SizeFixer (someone did it for me) at normal setting (because it
can't read the Casio exif for advanced mode)- very much like 1), but
with worse jaggies on diagonals.

7) Photozoom Pro - Ran demo with S-Spline (supposedly the best) at
default settings.  Result: Terrible.  Looks horribly overprocessed and
cartoonish.  I didn't try any of the other algorithms.

8) Jack Flesher's PS-CS/bicubic workflow, as described at this link
(thanks to Carl Schofield for the tip):

  http://www.outbackphoto.com/workflow/wf_60/essay.html

It goes like this:
a) do all work to get the image ready to print, including sharpening,
 at native resolution
b) upsize with bicubic smoother to 20% past target resolution
c) add more sharpening
d) downsize with bicubic sharper to target resolution

The result was much better than my straight bicubic workflow, in these
ways:
- sharper, with fewer sharpening artifacts
- better shadow separation and low end contrast
- over all look was sharper & contrastier with better shadow detail

I could not match the results with my normal methods.  My workflow has
been this:
- convert to grayscale
- then upsize with bicubic
- then do the work

Then I did some experiments, and found that the real difference is in
doing the work, including sharpening, before upsizing.  When I changed
my workflow to this:
- convert to grayscale
- do the work, including sharpening
- then upsize with bicubic
- add a bit more sharpening

...the result was nearly identical to the Flesher workflow.  That
print was still a tiny bit better.  I did some tests with the workflow
and the crucial step seems to be in the sharpening added between the
two resizings.  But it's real easy to add too much and it starts
looking overprocessed.  Too little and it doesn't look any better than
the straight bicubic step.  So it's playing right on the edge, looking
for the sweet spot.

Seems like the really big difference comes from doing all the resizing
after the work is done, rather than before it.


Regards,
Clayton


Info on black and white digital printing at    
http://www.cjcom.net/digiprnarts.htm