On 12/3/2011 4:50 AM, Chuck Norcutt wrote:
> Your struggles with the exposure foibles of the E-1 reminds me of an
> article I read recently chastising our camera designers for not making
> full use of the exposure measurement technology already available to
> them. It covered more ground than ETTR but, since ETTR is of special
> interest to me, it's what I paid most attention to. The author simply
> pointed out that any camera with live view could easily implement an
> ETTR exposure mode and guarantee that no highlights would be blown since
> it knows the exposure state of every pixel.
It seems to me that the author's view is simplistic. Implementing auto ETTR
in-camera is trickier than one might
imagine. Let me give examples:
1. When scanning film with VueScan, it is possible to do exactly what he
proposes. Simply set White Point to 0 and VS
will expose/process so that the highest brightness in any channel of any pixel
is defined as the top value in the scan.
So nothing is clipped. (Same choices at the bottom)
With some frames, this works just fine. With frames that have just a few points
exposed much above the rest, it tends to
underexpose the image as a whole, sometimes very seriously so. A WP value of
0.01% is often enough that overall exposure
is good and there is no visible clipping. I've never seen any visible clipping
at 0.1% or even higher.
2. When editing an image in PS that has small, specular highlights and an
overall histogram that just kisses the top,
select a small area with specular highlights and check its histogram. Oops,
clipped values! This is essentially the same
problem as in example 1.
3. When editing subjects like water rushing over rocks in direct sun, I usually
work to avoid blown highlights. Quite
often, once I've carefully dealt with them, and have a 'perfect' looking histo,
I realize that the image now looks less
appealing than at an earlier stage. The truth seem to me to be that for some
subjects/circumstances, a modest amount of
clipping in the right places simply looks better.
Now some bits of information:
1. Since the beginning of photo processing for consumers, prints were made with
intentionally contrasty midtones and the
highlights and shadows allowed to clip. The reasons are simple.
a. The average camera user/print consumer preferred punchy snaps to flat
ones with broader dynamic range.
b. Even with the latest computerizes processing, it wasn't practical to
custom tune individual images to the needs
of the subject.
2. Things haven't changed that much with digital. The vast majority of cameras
sold, and thus those from most of the
profits are made, are judged by the images that show up on their LCDs or on
prints from automated machines. People go
into the shop, snap a few shots on the demo and judge the camera based on what
they see on the LCD. Or they look at
sample images on the web.
There has been improvement. Quite a few cameras now include algorithms to
evaluate characteristics of subjects to select
exposure and internal processing parameters. These filters are based on
analysis of at least thousands of shots of
common subjects. The smart auto mode on the Sammy WB650 is far better for a
shooter like Carol than simple auto the Fuji
F10 she was using before. I, on the other hand, get about the same sort of
results with either F30 or Sammy.
So here's the dilemma for camera makers who might want to include auto ETTR:
On one hand, their ability to produce images that will please the buyers of the
majority of their cameras without
clipping highlights a lot of the time is still limited. ETTR isn't practical,
or at least doesn't add perceived value to
consumers.
On the other, the technically sophisticated photographer is likely to find any
particular implementation of auto ETTR
parameters to fall short with too many subjects:
1. Set WP to zero, and too many shots will be seriously underexposed. Then when
the midtones, and especially the
shadows, are pulled up to balance the image, noise rears its ugly head. Unhappy
users!
2. Set the WP too high, and another (probably overlapping) bunch of users will
be unhappy with the clipping.
There are also technical limitations:
1. The vast majority of mirrorless cameras using live view probably don't have
the processing power to implement auto
ETTR as proposed without slowing them down to unusable levels of performance.
Even high end compacts and most ILCs are
not rockets in performance already. Add the requirement to evaluate 10-20
million individual pixels in real time to set
exposure - not with current price/technology.
2. Live view on DSLRs is not the primary mode of use for most shots, for most
people. I love live view, but it still is
probably used for no more than 10% of my shots. And it is, of course,
impossible to do live, auto ETTR with the mirror
down. (Maybe with an E-300/330 hooked up to an external computer?)
3. For a great deal of live view uses, live histograms, bracketing, chimping,
etc. are quite practical ways of
controlling ETTR.
When will we see auto ETTR? When a significant maker adds it as a feature and
promotes it as their unique advantage. If
it then catches sales, everybody will add it, whether it is actually useful
and/or works worth a darn or not.
What are camera makers doing to address the need to minimize clipping today?
1. "Smart" scene modes are working their way up into lower level DSLRs. I've
never used them, but I think the 60D has
such a thing. I believe the same is true for other makers.
2. I haven't kept up with all the endless DSLRs out there. I do know the D60
has a mode that modifies the shoulder of
its response curve to have a much shallower roll off - and up to a higher level
of subject brightness, than the standard
setting. In effect, it compresses the highest tones, rather than clip the very
highest. Given the very large number of
value steps available at the top of a 14 bit image, this allows much greater
recovery of highlight details in post.
The price is a minimum ISO of 200, and the slightly higher shadow noise of 200
vs 100. Overall, I have found it pretty
effective in extending practical DR and guarding against unintentional lost
highlights.
It directly addresses the issue you raise with the 5D, of setting ETTR using
the histos, and still having some clipping.
The 5D is a great camera, but pretty old technology in some areas.
Technically Wordy 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/
|