Moose, just a query about deconvolution sharpening:
Does the algorithm for deconvolution take into account the lens used ?
I would assume that the way the lens "convolutes" the image would vary
considerably with lens type, e.g. a simple tessar design compared with this
new m.zuiko 12-100mm.
...Wayne
-----Original Message-----
From: olympus
[mailto:olympus-bounces+wayne.harridge=structuregraphs.com@xxxxxxxxxxxxxxxxx
] On Behalf Of Moose
Sent: Thursday, 15 September 2016 5:51 AM
To: Olympus Camera Discussion <olympus@xxxxxxxxxxxxxxxxx>
Subject: Re: [OM] Megapixels and detail in images
On 9/14/2016 2:15 AM, bj@xxxxxxxxxxxxxxx wrote:
> I have to confess that this issue puzzles me. Yes, I know it is not
> the only one ...
>
> For some time I have been somewhat disappointed by the amount of image
> detail provided by OM digital cameras.
It's hard to know just what you find disappointing without more description
and/or examples.
> The best two I have are very close in performance as far as I can
> tell; these are the E-3 and the DE M5.
It sounds like you are treating MP numbers as linear indicators of
resolution. This is not the case. MP is an area measure, while resolution is
a linear measure. The increase from 10 to 16 MP might lead some to expect a
60% increase.
But look at the dimensions.
E-3 = 3648 x 2736, E-M5 = 4608 x 3456
So nominal linear resolution has increased by 26%, not 60%. I say nominal
because sensor systems aren't all created equal. For example, when I moved
up from a Canon 300D with APS-C size sensor to a 5D with full frame sensor,
the number of pixels on the APS-C portion of the 5d sensor was less than
that of the 300D. Yet, in careful testing, I found that the 5D resolved
slightly MORE details than the 300D. [I see AG has had a similar experience
with E-3 and L1]
Some experts, or at least folks with lots of experience testing cameras, say
that any increase under about 20% is unlikely to provide more visible
detail. I could go into more detail as to why this may be so, but the point
is that it has been my experience as well. Because of the need to demosaic
the RGB patterns of sensors, most color sensors are only about 50% efficient
at rendering the resolution that the lens delivers to them. Some of the
difference is in the quality of the sensor system, some, for those who shoot
RAW, in the demosaicing algorithms in the Raw converters.
So, It's not really surprising that you see little difference between these
two camera bodies in resolution.
I also need to refer back to your last post of images. You shot at f22, and
were pleased with the DoF, and apparently with the resolution? This is a
real problem for resolution. F22 is getting deep into diffraction territory
on a 4/3 size sensor.
If you were to look at a test like this of the 12-50 you are using (and that
I use extensively), you might be forgiven for thinking that stopping done
has little effect on resolution, especially at the long end.
<http://www.photozone.de/m43/827-olympus1250f3563ez?start=1> But if you are
paying attention, you'll notice that the charts don't go past f11, for the
reason that things get a lot worse. Then notice the text "Since the lens is
quite slow, it basically runs immediately into diffraction effects so
stopping down has only a slightly positive effect on the outer image region
whereas the center quality decreases."
So what happens with a fast lens? Here's the excellent M.Z 75/1.8.
<http://www.photozone.de/m43/830-oly75f18?start=1>
Resolution is wonderful at larger apertures, the best they had tested at the
time. BUT, by f8, it's essentially the same as the 12-50 @ f8. Again,
"Diffraction has a limiting effect from f/5.6 onward . . ." Diffraction is
the great leveler that can bring the finest lens to its knees.
While the sensor is 16 MP, you are choking it down to 4 MP or less. Chuck
guesses under 2 MP. If your sensor were 36 MP, you would still get little
resolution, because the lens just can't deliver it, a fact of optical
physics. You need to avoid carrying over ideas of usable apertures directly
from FF to 4/3. While f22 isn't too very bad for some uses where lots of DoF
is needed on FF, and display size is modest, it just whacks resolution on
the smaller sensor.
AG says that it isn't quite this bad, that in practice, you get 1-2 stops
better than theory says. He's proposing that the above sort of tests
overstate the blurring from diffraction. But what he's saying is that f8
isn't too bad, not that
f22 isn't awful.
If you are shooting 4/3 @ f22 and expecting good resolution, you are bound
to be disappointed.
Beyond all that, firmware and software DO make a difference. I recall that
you mostly shoot JPEGs, thinking they penalize little but save disk space
(really, really inexpensive disk space). The practical truth is that
shooting only JPEG simply throws away some resolution. Sure, there's some
combination of camera sensor system, lens and subject, somewhere, where it
doesn't, but in ordinary use, it just does.
Then again, you use only free software. Raw conversion by FastStone is
second rate, that with Oly Viewer pretty good, but not excellent. AG
casually mentions when talking about diffraction effects that it's not as
bad at portrayed in part because of "because of sharpening tools available".
I also find this to be true. To MikeG's puzzlement, in spite of the
theory/math, deconvolution processing ("sharpening") does somehow recover
some of the resolution lost to diffraction. So
f8 can be more like f5.6 or so. But he is not referring to simple USM
sharpening, such as that used by FS. He's talking about tools you have to
pay for, and learn to use properly.
I also know that other aspects of your technique are decreasing detail
resolution. You steadfastly refusal to use negative EV compensation in
bright, high contrast situations, witness the extensive blown highlights in
your recent daffodil image. I've written before, more than once, explaining
how this eliminates detail - reduces resolution, in highlight areas. I'm not
going to repeat the details yet again here, but it's true.
In summary, at every step, from choice of aperture through post processing,
you make decisions and take actions that reduce resolution - then complain
about the results. It's like keeping a hedge trimmed carefully to three
feet, then complaining that it gives no privacy. If the folks rebuilding
your house were to be as sloppy in their work, you would be livid.
> I've just taken delivery of John shaw's book "Digital Nature
> Photography" in which he points out that he uses exclusively Nikon gear
and the Nikon D800e has 36.30 Megapixels.
First of all, you aren't even coming close to the resolution the camera you
have can deliver.
Second, there is a price to pay for very high resolution on a 24x36 mm
sensor. Nikon does this by eliminating the AA filter effect on the 800e, and
eliminating it entirely on its replacement, the 810. The price is moire
effects, including very bad false colors in fine repeating patterns.
A Nikon 810 is far, far more expensive than proper Raw conversion and
editing software, and requires more attention to details of use/technique
than you are apparently willing to expend on your excellent, 16 MP camera.
As to what the camera is capable of, back in April 2012, when I was
contemplating buying the same camera you are complaining about today, and
that I did buy that Aug., TOP offered a sale of Ctein 15x20" prints made
from the very first µ4/3 camera, the 12 MP E-P1. I bought one, and can say
that it's much more impressive than the web posting shows.
Detail resolution is spectacular.
<http://theonlinephotographer.typepad.com/the_online_photographer/2012/04/ho
w-i-made-the-1995-print.html>
As the first comment on the blog points out, this is a careful effort by a
PS expert and master printer. I'm not suggesting that you should try to
match that. I am suggesting that it shows that our cameras and lenses are
generally capable of greater visible detail than we manage to get out of
them. Blaming it on the camera is usually wrong, a convenient way to excuse
what I've done with it.
> So I thought I'd see how that impressive figure compares with the M5's
16.1 MP.
>
> The first site . . .
> As I recall, the Nikon had no visible advantage.
That's silly, unless the test technique is poor - or - display size is small
- or - they are looking at prints, even quite large ones. Tests of the
effectively about 40 MP High Res Mode of the E-M5 II show it to be in some
ways superior viewed at 100% to the 810, as in many examples here.
<http://www.imaging-resource.com/PRODS/olympus-e-m5-ii/olympus-e-m5-iiTECH2.
HTM>
I found the same sort of thing, where the slightly higher resolution of the
810 is offset by loads of false color.
<http://www.moosemystic.net/Gallery/tech/X-T1_E-M1_D750/FON.htm>
I mention the HR Mode not to encourage you to buy an E-M5 II, but to draw a
parallel. As I recently wrote, the difference between the 16 MP of the E-M5
(and 5 II) and the HR mode of the II, while apparent when pixel peeping,
does not make more than very subtle visible difference in large prints. I
believe the same would be true of the 810.
> There are many sites about he M5, but few making the comparison I want.
I assume this is for the simple reason that's it's an unfair comparison,
that doesn't show anything useful. The Nikon 810, and Canon eq. do indeed
resolve more detail then that of which the E-M5 is capable. BUT, the E-M5 is
capable of far more than you are getting out of it.
Moose D'Opinion
> It might have
> to be phase 10 !! :-)
>
> Brian
>
>
--
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/
--
_________________________________________________________________
Options: http://lists.thomasclausen.net/mailman/listinfo/olympus
Archives: http://lists.thomasclausen.net/mailman/private/olympus/
Themed Olympus Photo Exhibition: http://www.tope.nl/
|