On 1/24/2017 3:37 PM, Mike Gordon via olympus wrote:
Subtlety Moose writes:
<<<I don't know if it's sensor limitations, lens (probably not) or imperfect
X-Trans sensor demosaicing, but there's a point in the finest feathery parts where
details just sort of mush away.
I agree with the Moose conjecture that much of the fault is the demosaicing
algorithm.
I know I harped on this before.
CFA's have trade offs by nature. Xtrans colored filter array is designed so the
"green" raw
channel has more coverage but less positional symmetry than in a Bayer sensor
and as a direct result there is less high-frequency luminance
"miscalculations" in the interpolation, giving lower noise in raw conversion.
The noise Fuji made was about defeating the moiré problems of Bayer arrays. This could be interesting, as all the higher
res sensors are starting to bump up against the Scylla and Charybdis of AA filters that defeat the purpose vs. moiré
problems, including wild false colors.
Looking closely at the sample images from the new, 20 MP µ4/3 sensor, I can see moiré increases over 16 MP. Subtle as
yet, no problem, but indicators of troubles to come if they go higher. I wouldn't be surprised if that's why they didn't
go to 24 MP. Oly particularly has to stay away from AA filters, so their HR Mode will work.
If Fuji can go to say 32 MP, without AA filter or moiré, the increased resolution might take care of the smeary effect
on resolution, giving the eq. of a nice, clean 24 MP.
The larger B to B and R to R distances necessitate a large chroma smoothing radius resulting in "watercolor" effects
or artifacts- some converters seem to smudge detail in the process as well -balance
What if there is, in fact, just not enough information in the R & B channels to get what we would consider full
resolution for 24 MP, no matter how is it processed? The fact is that this level doesn't matter for anything but
gigantic prints.
I sent some of my HR Mode images to Ctein, noting how much more pixel level detail there is in an HR file downsampled to
the 16 MP of the sensor. He wasn't surprised, saying "Keep in mind that a conventional Bayer array photograph is only
about 50% efficient in terms of spatial frequencies-- that is, you get about eight megapixels worth of actual fine
detail out of a 16 megapixel file."
I sort of took his word for it, as he has more experience and understands the math - and it makes sense - and esp as it
agrees with my empirical results. The difference between a straight 16 MP image* and the same size downsampled from the
HR ORF, is almost startling to someone used to pixel peeping the 16 MP files.
With the Fuji R & B sensels spaced twice as far apart as in a Bayer array, might we be seeing the limit to color
resolution?
I know Fuji doesn't have IBIS, but I wonder what would happen if they built in the ability to take two shots in a row, a
sensel apart, perhaps diagonally, It still wouldn't cover every pixel location with a sensel of each color, but would
provide much more chroma resolution info for a tricky demosaicing algorithm.
seems different with different converters. The ACR in PS and LR used to be
plain lousy with RAF files, but improved , if suboptimal now. Accuraw allows
one to adjust the balance of artifacts and more resolution vs watercolor
effects--though have never used it.
http://www.dmcgaughey.com/2012/12/31/fuji-x-trans-raw-conversion-accuraw-takes-a-swing/
Please see some of these early comparisons.
Is a more perfect solution theoretical possible? --- Probably yes.
Or, as I speculate above, might the choice be what Accuraw offers, a trade-off,
depending on subject?
Someone with less rusty math might suggest a POCS algorithm or using a more
advanced multi-pass interpolation algorithm with best fit iteration for a
layout that is inherently asymmetric.
The penalty is the very resource intensive nature of these techniques and not
sure if people would want to wait several minutes for one conversion.
Is it real? Or is it CSI "enhancement"?
Between Pixels Moose
* Oly actually saves one of these, in addition to the ORF and JPEG HR versions, when you take an HR shot. They have a
file extension of .ORI. Rename or put in a different directory and change the extension to .ORF, and they are just what
would have been taken in normal mode. I assume the first exposure of the eight used for HR.
--
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/
|