On 5/19/2016 4:34 AM, bj@xxxxxxxxxxxxxxx wrote:
Moose wrote a whole lot which is beyond me responding to right now,
except for these two lines.
. . .
Please note I did not complain about
leaf colour - luminosity. Resolution was the topic.
Quoting myself, which you appear to have ignored:
'"Why is he beating this dead horse of highlights, again?", you may well ask. "Because it affects detail resolution, the
subject of your post!", I holler back. I've explained how/why many times, both in response to your posted images and
those of others, and it seems to done little good (except for Jim). For now, IT JUST DOES.'
Yes, it affects colour, yes it affects other aspects of the image - but I was staying strictly on topic - resolution.
Here is a 100% sample of a shot I took yesterday of a California Poppy.
<http://www.moosemystic.net/Gallery/tech/Overexposure&Detail/PoppyDetail.htm>
See all that fine surface detail? Then look at the second version. Here, I've done something very like what Oly's JPEG
engine does, compressed highlight detail, without clipping. See how the details start to disappear, esp. in the brighter
areas?
The third version shows what happens as compression isn't possible, either because the original pixel is already blown,
or because of the overall compromises the algorithm makes. Nearly all the detail is simple gone (And, the colour gone
far off.)
COMPRESSION AND CLIPPING DEGRADE RESOLUTION!
What ever you may think, whatever theories you may come up with, theory and speculation must give way to experimental
results. (Correct theory does, in fact, agree with the above conclusion.)
This is one reason not to rely on JPEGs. The data from the sensor is 12 bit*, so there are 4096 possible brightness
levels for each color channel. a JPEG is 8 bits, 256 levels. The camera's JPEG engine has to reduce tonal resolution by
a factor of 16, in a split second. It can't just squish everything down proportionately; the result looks very wrong to
human vision. Worse, it is often expected to increase contrast, esp. mid-tone contrast. The only possible solutions are
to chop off and/or compress the dynamic range of highlights and shadows.**
Compressing the top highlights is a better solution than chopping them off, as it allows the kind of partial correction
I illustrated in these samples from your posted images. It is not, however, as good as shooting Raw. There is a loss of
data in the compression, and the uncompressed version is always a little off what would have been possible.
As you should be able to see, chopping off the top is even worse, but that's what you are doing when you don't use EV
compensation for this kind of shot. A pixel runs out of electronic room to record the brightness, and it fails to record
tonal detail needed for detail resolution and color accuracy. And sometimes JPEG engines do the same.
======================
Lens filter. On
for lens protection. Clearly, if I am to be super-fussy, I need to take
the risk of removing filter for the duration of shots.
Or, use a lens cap, instead.
Or, test the filter by a careful shot with and another otherwise identical, without. If they look the same @ 100%, go
ahead and use the filter. My comment was meant to point out that a complaint about resolution when all steps to
eliminate causes extraneous to lens and camera is potentially meaningless. You have nicely illustrated this. You didn't
check everything, and had the wrong camera setting.
* I think. If it's 14, that just makes the problem worse.
** This is nowhere near a new problem. Automated color processing and printing, from drug store to high end lab has had
the same problem, and used similar solutions forever. You would be amazed to see the detail in the color negs that
didn't make it into the print.
--
What if the Hokey Pokey *IS* what it's all about?
|