Chuck Norcutt wrote:
> I see what appears to me to be improved sharpness and contrast but I
> wouldn't call it detail.
OK, on one thing I think we can agree to agree. There is no more actual
detail in the up-sampled image than in the original. Despite what they
do on TV forensics shows, data that isn't there can't be "extracted" or
"enhanced".
Then we may be able to agree on another thing. Data can be in the image,
but not apparent to the human visual system. Reorganization of tonal
data can make things visible that weren't visible in the original. I
hope you have seen enough blue sky & clouds appearing out of
undifferentiated white and detail rising out of shadows in examples I've
posted to believe that. Like the many details that are invisible or only
hinted at in the original here and clear in the adjusted version
<http://galleries.moosemystic.net/Summita/pages/DSCF1140a.htm>.
I then propose that the FM plug-ins, and especially the Resize Pro, make
detail that was in the image, but unseen, visible and/or more clearly
visible than do the bicubic versions.
> And I'm not at all sure that 200% pixel peeping bears any relationship to
> prints.
>
Careful now, this is 100% pixel peeping at up-sampled images. A
significant difference.
> Why haven't you compared them on paper at largish print sizes?
>
1, Lazy.
2. I haven't printed anything bigger than 8x11 in ages, and haven't
needed to uprez.
3. Most of my images are viewed on screen, not on paper.
4. Did I mention lazy?
And - after I did this test, I read Harald Johnson's book on digital
printing, where he convinced me that the printer drivers are going to do
a better job of matching the available image pixels to the actual
hardware than any up-sampling to match my imagined version of the best
dpi for my printer. He claims his conclusion is the result of extensive
experience and testing. I choose to believe him for a few reasons:
1. I'm happy with the printed results I'm getting.
2. The other things that he wrote that I tried turned out to work.
3. Did I mention a personal aversion to work that doesn't accomplish
anything?
Remember, I wasn't trying to sell anything in this thread, only to show
how my test results with a full color image of a natural subject differ
from those using a B&W bar chart in the link posted to a web page on
Paul Butzi's site claiming that stair interpolation is worse than
simple, one step bicubic smoother.
I am not advocating that anyone buy anything. In fact, I would suggest
first simply leaving the image at the pixel size it is, setting the
printed output size in the printer driver and letting the calculated dpi
fall where it may.
Moose
==============================================
List usage info: http://www.zuikoholic.com
List nannies: olympusadmin@xxxxxxxxxx
==============================================
|