> The default setting of RSE has very strong sharpening. Ok, at
> the same time they keep the sharpen artifacts in good
> control. Even I set the output -50 (minimum position) but
> there is still some sharpening done.
I did notice that. It came to light on my first test image when
I was cranking settings to extremes to figure out what all they
did. Lo and behold I was getting halos when I thought the
sharpening was turned off. That's when I found the default
camera settings and wrestled the sharpening there. That took
care of the halos no matter how much I destroyed the image after
that.
There are physical limits to how much detail you can wrangle out
of a RAW file. You can't change the law of physics. However
each RAW converter does the demoisiac ever so slightly different
and deals with the AA filter in unique ways. I'll be doing some
test chart stuff this weekend, hopefully, and get a handle on
just what and where these differences are.
I have one real problem image taken with the A1 (trainyard and
capital building) that has a lot of nasty alaising artifacts in
it. Due to the converging lines and tight repeating patterns,
this image is a worse-case example of a "res chart in the wild".
(a film version of the image appeared in TOPE 20). My first
attempt at using RSE on the image resulted in a tremendous
increase in resolution (the bugaboo of the Ax cameras) but
didn't help the artifacts any. No better, but no worse either.
Converted in the Minolta software the image is soft and needs
three-pass sharpening to be usable, but the edges are straight
and the pixels line up and no PF is visible. The artifacts, when
you can see them, are rainbow colored. When converted in
Elements 3.0, the image is sharper but PF shows up and the
artifacts are rainbow colored. Converted in VueScan there are
zigzags as the demoisiac algorithm appears to be off by 1/2 to 1
pixel every other line. The artifacts (as recorded by the
sensor) are more visible and can only be disguised by adding HSM
noise to the image. In RSE, the pixels are lined up perfectly,
PF is pretty much non-existent and there are details showing up
that I've never seen before--not even in the Velvia version!
The artifacts (as recorded by the sensor) are still there, but
strangly different as they are more monochromatic rather than
rainbow colored. What is most interesting, is the sensor noise
in the RSE images is present, but much tighter and
monochromatic. A slight bit of luminous noise reduction softens
that noise without trashing too much detail. In all other
converters, the noise is colored and responds to color noise
reduction instead of luminous noise reduction.
Now, how does this translate to E-1 images? Not sure yet. I
love the colors that the Olympus software gives me, but the
images seem to be resolution limited by the AA filter. As I was
converting 175 images last night (all of which had already been
converted using Olympus Viewer) I noticed on a couple specific
images the lack or change in noise. Just like the A1 pictures,
the noise appeared to be more luminous (gritty) than color in
the RSE images.
Is this an improvement? Well, for landscape images, I do
believe so. However, earlier in the evening, while I was still
learning the ins and outs of the program and processed a couple
wedding pictures. The gain in texture details in fabric and
hair was offset by a marked change in skintones. The change in
the noise pattern from color to luminous caused a cadaver look
to erupt from the skin. Instead of a nice healthy pinkish glow,
the skin turned more gray. When viewed at 100%, an E-1 image
(when converted in-camera or in Viewer/Studio) tends to have a
colorful splotchy noise pattern that is "film like". Seriously,
take a close look at Portra films and you'll see the dye-clouds
that give the same splotchy noise. The RSE version seemed to
lose that colorful splotchy look and replaces it with brightness
noise instead.
The only comparison I can think of is Ektachrome 400. The grain
is very gritty, but not colored. However, Kodak gold 400 has
colorful grain, but doesn't have the same gritty look.
Much more testing to be done. I'm hoping that this discussion
sparks interest in other list-members to also do testing and
contribute to the joint learning session which is the OM list.
AG
__________________________________
Yahoo! FareChase: Search multiple travel sites in one click.
http://farechase.yahoo.com
==============================================
List usage info: http://www.zuikoholic.com
List nannies: olympusadmin@xxxxxxxxxx
==============================================
|