Fuzzy Moose wrote:
> First, you have gone to generalizations, away from the specific lens and
> photo Mike and I have talked about. Fine in a
> way, but ignoring some of the oddities in it that I don't believe can all be
> dismissed this way.
You are absolutely right that I'm generalizing. While this set of
photos may point out issues with a specific lens/camera combination,
there are certain imaging characteristics which bias the results from
one lens and/or camera to another. The bulk of my digital imaging
experience has been with the A1, E-1, E-3, L1, and 6D. Recently, I've
added the GX85, E-300, E-400 and A7ii to the mix. The generalization
holds together pretty well when dealing with high-frequency,
low-contrast items. It is also true that lenses are a major
contributing factor. Some lenses are far better with micro-contrast
than others. The PanaLeica 14-50 Four-Thirds zoom is one such lens.
The subtlety between similar tones and colors is enhanced by this
lens. Of the legacy OM Zuikos, the 50/1.4 (final series) has more of
this trait than the 50/1.8 (final series). And this is evident on the
A7ii. With the 6D, there was no visible difference as the sensor
successfully managed to mask such subtleties.
> I believe most of the problem we actually see is a result of the Bayer array
> demosaicing. What you are talking about is
> likely true, but masked by the worse effects of the Bayer system.
> Tack a pelt on a wall, back off to the point where the mush happens. Then use
> an OM-D body with HR mode. Shoot in HR,
> then downsample to sensor size, compare it to the regular shot, and there's
> all sorts of better detail. Ctein figures
> that about half of the inherent sensor resolution is lost to Bayer
> I mention only Oly because I don't know which others do HR and how they do
> it. Oly samples every sensel location with
> each color sensel, so color is direct, not calculated. The difference is not
I believe this is precisely related to the micro-contrast issue and
not reduced sensor resolution as implied by Ctein. The older
converters would work with and average three or four source pixels
together to determine the correct RGB value for the converted pixel.
But now, the algorithms are far more complex and have the uncanny
ability to get down pretty much to a 1:1 source pixel to converted
pixel. My direct experience with the "oversampling" HR mode images is
limited, but I've looked at samples on-line. What I'm seeing is less
about raw resolution and more about more pixel to pixel contrast.
Major contrast items are easily captured by pretty much every imaging
systems, but capturing and representing pixel to pixel contrast of
very slight gradient or color shifts (as seen in flowers, leaves, and
Themed Olympus Photo Exhibition: http://www.tope.nl/