Hmmm,
Excellent methodology in the test. and a bit surprising but not
shocking. One possible other variable is noise. Deconvolution hates
it. I think the GM-1 files may have a bit more
than OM 5DII and so far have used "Prime" noise reduction first if
pushing the global DOP lens softness slider past default--even at base
ISO. That may eliminate some artifacts. I don't know if you made
virtual copies and tried
bumping up the "details" slider. One hypothesis is that the "global"
adjust pixel radius and "details" iterations DOP deconvolution may
leave the central part of the image almost untouched if it is tack
sharp.
I think you are also spot on about using deconvolution after DOP or
second round with FM-- After about 20 experiments I think what happens
is the artifacts from the first round are accentuated--(?ringing)
and it often works best to be a tad more aggressive once. One can use
the "microcontrast slider in DOP but I think I like the control in PS
USM a bit better.
I agree that if the overall image detail in most areas is not improved
with DOP, the highlight testing is immaterial. As long as you
downloaded the free month long trial, might be worth trying the prime
noise reduction on a few higher ISO images. Theoretically removing
noise in the RAW stage could be superior.
I am glad to have DXO in the tool box for the 10% of images I use it
for--better with skin tones with the 5DII and much quicker for me with
high ISO images and usually better with predominant flash shots with
people, IMO. Also color is better when purple should be accurate
especially with skin tones in the same image. That is VERY cam profile
specific. I still don't know about the GM-1. I also don't like being
limited
by the metadata for the geometric distortion correction. Thus far the
ACR choice has been just fine,
but sometimes it may not be. I like the way the adjustments wit the
lens modules works in ACR much better for Canyon.
I was hoping DOP would now edge out ACR (knew it was close) and you
would figure out all this stuff especially with the GM-1 and I all
I had to do
was pay attention on the list. :-) It will correct CA with Oly lenses
on Panny bodies too of course
Mike
On 7/11/2014 8:50 PM, Mike Gordon via olympus wrote:
FSM writes:
In my long ago test, DxO's deconvolution specifically developed
foreach
lens couldn't beat Focus Magic on the lens
I tried, and a couple of my most used lenses didn't have DxO
profiles.
That test was interesting but not exactly what might determine DXO
Optics Pro
(DOP) utility in your workflow, IMO
anyway. Several versions ago they recognized the elephant in the
room and
instead of having to save a 16 bit tiff to
disk then open it in PS, you can just do what you need to do in DOP
and export
directly--quick.
Yes, I've just done that. BUT, it must demonstrate superiority to
alternative
Raw conversion to justify buying/using it.
Also you used the default slider positions--asking alot of defaults.
I have just done my first comparison. One of my biggest interests is in
shots
taken with the 75-300 @ 300 mm.
Traditionally, the long end of long zooms is the softest and I use that
a lot.
I chose a shot at quite a distance, at least a mile, without too much
depth, so
everything should be in focus, and lots
of complex, subtle detail, rather the antithesis of test charts.
DxO applied much more linear distortion correction than ACR. As ACR is
supposed
to use the correction data from the lens
itself, I'm a little at a loss as to why the DxO correction should be
so
different. Visually, it seems overdone, and I
only shot this a month ago. If I turn off Distortion Correction, the
result is
almost identical to ACR.
With lenses that have weaker areas, the deconvolution used gently can
even up
the performance and then use FM in PS in
a layer(s) as you do for fine tuning---FM and DOP are not mutually
exclusive.
No, of course not. However, in this example, that's not an effective
option. The
default "DxO Lens Softness" control
settings are, at first glance, a nice improvement over unsharpened ACR
output.
It's pretty clearly a combination of
detail enhancement through deconvolution and/or USM and LCE.
When compared to ACR with no sharpening and FM with a little added LCE,
the DxO
version clearly is slightly inferior,
i.e. slightly more subtle detail is clearly visible in the adjusted ACR
version.
This is true center, partway out and
edge, with no variation I can see.
I tried going further with the DxO version, as you suggest, but even a
one pixel
application of FM leads to obvious
artifacts that make things worse. That led me to look closer, and FM is
simply
making very subtle artifacts more
visible. The unmodified DxO conversion has artifacts that aren't in the
default
ACR, nor in the adjusted versions.
These artifacts are pretty subtle stuff. I'm not sure the unobsessive
would
notice what the Eagle Eye sees at 100% on a
24" monitor.
Adjustments of the DxO Lens Softness control don't help, either
reducing detail
or increasing artifacts.
In any case, with this particular lens, on an E-M5, at this focal
length, ACR,
FM radius 3 and a little LCE clearly show
more true image detail than I can get out of DxO (DOP). FM with radius
4 is even
better, but right on the edge of
overdone. I'd probably use 3 with a partially opaque layer of 4 above
for my
best result.
Also you are not bound by the geometric correction metadata and can
pick how
much to torture pixels and change FOV.
In this first case, as above, I think the default lens correction is
probably
right on. It should be, at the long end of
a tele zoom. Perhaps I'll see what happens at the short end and close
focus of
the 12-32.
I had meant to do several different conversions with a few of
Marnie's
Catalonia images last weekend with ACR/DOP with
comparisons but the limited image processing time was inadvertently
consumed
by the B'fly shots.
That seems a worthwhile diversion of resources. :-)
I still don't have the color handling figured totally with GM-1
shots and
DOP.
I did nothing with color. The test shot of complex granite is close to
B&W. AG
finds Capture One better for the
subtleties of portraiture, but I do so little of that that it's a non
issue. DPR
saw no meaningful difference in color
between ACR and CO, which doesn't prove much.
Oh, the highlight treatment improved with the 9 version ago as well,
but not
quite up to ACR.
Nor did I look into this. The test shot, intentionally, is one of
limited DR,
reaching neither end of the histogram. I
wanted no other effects interacting with what I was looking at.
...
I truly do think it time to reevaluate--perhaps a couple images that
need the
corners sharp with a lens that also has
a bunch of aberrations.
Well, I'm not looking for lenses to torture test anything, just the
lenses I use
most. I don't think they mostly qualify
as particularly soft in the corners nor full of aberrations. They may
or may not
"Draw" as some might prefer, but are
good in measurable IQ.
Would just use DOP or ACR as the front end and do your skillful final
processing in PS for both.
Well, that was the idea, but so far, there is less than no point. FM
and LCE are
Actions in PS for me, so application is
no extra time/effort and the results are superior.
No mad rush really--the sales seem to be more frequent than the
Canyon lens
rebate times. One can be guaranteed there
will be another in just a bit.
Good to know. I may manage one or two more tests before we are on the
road, but
can't see buying just now, based on what
I've seen so far.
Critical Edges Moose
--
What if the Hokey Pokey *IS* what it's all about?
--
_________________________________________________________________
Options: http://lists.thomasclausen.net/mailman/listinfo/olympus
Archives: http://lists.thomasclausen.net/mailman/private/olympus/
Themed Olympus Photo Exhibition: http://www.tope.nl/
|