From: Floyd L. Davidson on
"DRS" <drs(a)removethis.ihug.com.au> wrote:
>"Floyd L. Davidson" <floyd(a)apaflo.com> wrote in message
>news:87eipeqnt0.fld(a)apaflo.com
>> "DRS" <drs(a)removethis.ihug.com.au> wrote:
>
>[...]
>
>>> That is not in dispute. If you payed attention you'd have noticed
>>> that in that very paragraph I said, "Since the exposure has not
>>> changed..." The fact that the exposure is the same is necessary to
>>> the point being made, which is that changing one and only one
>>> variable, the contrast setting, changes the accuracy of the
>>> histogram upon which decisions about exposure are *subsequently*
>>> made.
>>
>> You can't have it both ways. First you say that the
>> contrast setting moves the histogram's indication of
>> exposure, and now you say it does not.
>
>I have always said it changes the indication of exposure. It is you who
>made the false claim that I and others were saying it had anything to do
>with changing the actual exposure, which we have repeatedly denied.
>
>> All five of those histograms show *exactly* the same
>> exposure, even though they have 5 different contrast
>> settings.

*All* *five* *show* *the* *exact* *same* *exposure*.

Not one of them shows any 1 f/stop difference as you
claim.

The only thing that moves 1 f/stop are the tall peaks in
the graph that have *nothing* to do with exposure. It
is well to the left of the right edge, and it is that
right edge that indicates exposure.

>And because of the different contrast settings the 5 histograms show the
>exposure differently. The exposure hasn't changed but the accuracy of its
>representation has. Which is all that has been claimed.

They *don't* show the *exposure* differently.

>> If that is the desired exposure, not one of those
>> histograms would suggest that the camera's exposure
>> should be changed to be more accurate.
>
>None of the 5 histograms indicate overexposure. What they do indicate, to
>different degrees of accuracy, is the room for exposure compensation. Only
>1 can be the most accurate and it is the -4 histogram.

All of them indicate there is no room for any
"compensation". Increasing the camera exposure will
result in clipping, and each of those histograms shows
that.

Oddly enough, the +4 histogram shows it the best!

>>> No, they don't. Blind Freddy can see the right-edges shift by at
>>> least 1 stop from -4 to +4, and probably more.
>>
>> You apparently are looking at that huge peak value to
>> the left of the right edge, as it is the only thing that
>> moves by "at least 1 stop". But that peak has *nothing*
>> to do with setting correct exposure. See the little
>> itty bitty value down in the lower right hand corner?
>>
>> That is the value of significance. That is the "right
>> edge".
>
>In this instance that indicates specular highlights, which as has been noted
>by several people, may be blown without spoiling the image. That is a
>choice by the photographer.

The tall peaks are *not* a "specular highlights", and
one look at the image should demonstrate that there are
no light sources or reflections in the image. Those
peaks are probably the whiter parts of the background,
but might be from one or more of the crayons too. It
isn't possible to determine from looking at the
histograms just where the actual values are that make up
either those peaks or the right edge of the histogram.
It does appear that the whitest part of the watch face,
around the 12 and 6, are probably where the right edge of
the histogram values are from.

Note that only in the +4 histogram are the values at the
right edge more than maybe 2 pixels high on that graph.
They are very hard to see because of the white portion
of the gray scale below the graph and the lack of
contrast with the gray background.

>>> The proof is in the pudding is in the eating.
>>
>> Cute. I'll settle for being logical.
>
>Good. I hope to see it soon.

With a bit of logic you'd have seen it already.

You know, logically if what you said made sense the
article would have discussed it in some way, but it does
not even hint at it. If changing the contrast displayed
a more accurate *exposure* (right edge location), why
didn't they point to it?

--
Floyd L. Davidson <http://www.apaflo.com/floyd_davidson>
Ukpeagvik (Barrow, Alaska) floyd(a)apaflo.com
From: Floyd L. Davidson on
Paul Furman <paul-@-edgehill.net> wrote:
>Floyd L. Davidson wrote:
>> That is exactly the point I *am* making. WB changes
>> the
>> histogram in a useful way. The Luminous-Landscape
>> article did not talk about WB at all. They discussed
>> changing the camera's setting for contrast, and that
>> simply does *nothing* useful for the histogram/exposure
>> issue.
>> And if you doubt that the contrast setting will not
>> change
>> the histogram as stated, *try* *it*. I gave an step by
>> step description of a very easy way to show exactly what
>> does and does not happen. Why argue from supposition when
>> you could actually learn something about photography...
>
>I am interested in learning. I tried fiddling with the camera contrast
>settings to compare histograms and the difference was very minor, at
>least for the conditions I tried just now. I was thinking of
>post-processing where increasing contrast blows highlights & blocks
>shadows. Maybe not the very final edge of 100% saturation but contrast
>adjustments can make a big difference in what appears to be blown or
>blocked. What you are describing sounds more like a gamma adjustment
>than contrast where only the middle part shifts.

"Gamma is equivalent to contrast. This can be observed
in traditional film curves, which are displayed on
logarithmic scales (typically, density (log10(absorbed
light) vs. log10(exposure)). Gamma is the average
slope of this curve (excluding the "toe" and
"shoulder" regions near the ends of the curve), i.e.,
the contrast."
http://www.imatest.com/docs/glossary.html
(The above is a web page owned by Norman Koren.)

"Gamma -- A way of representing the contrast of an
image, shown as the slope of a curve showing tones
from white to black."
http://www.hp.com/united-states/consumer/digital_photography/articles/scanner_glossary.html

That might help make things clearer as to why changing
contrast does not make the histogram any more or less
accurate for exposure.

When the histogram indicates precisely the correct
exposure, changing the gamma (contrast) moves the curve
between black and maximum white values, but it doesn't
change the value of either. Note too that it might
spread the histogram out from the center (if the gamma
curve is moved by picking a point at its center), or it
might move towards ether the right or the left (if the
gamma curve is moved from point closer to the ends
instead of at the center).

In any case, it changes the contrast but if there are
image values at the maximum white value they will not be
moved.

>An interesting related issue I don't understand is how the exposure
>slider works in Lightroom or ACR. I don't know how to duplicate that
>effect in photoshop with curves, levels, etc. Those all do like you
>describe, moving the middle parts of the histogram but there isn't an
>easy way I can see to shift the whole exposure. Hmm, the middle slider
>on levels comes close but still doesn't match the effect.

I don't use those programs so I can only provide general
instructions about how that is typically done. I'll
leave it to someone who can be more specific.

--
Floyd L. Davidson <http://www.apaflo.com/floyd_davidson>
Ukpeagvik (Barrow, Alaska) floyd(a)apaflo.com
From: Floyd L. Davidson on
Paul Furman <paul-@-edgehill.net> wrote:
>John Sheehy wrote:
>> Paul Furman wrote:
>>
>>> OK, the noise is more the reason but posterizing can
>>> be a problem in dark areas...
>> The only cameras I know of with even a hint of RAW
>> posterization are the Pentax K10D, which would profit
>> from 13 bits instead of 12 at ISO 100 (not for 200 or
>> higher), the Sony A900 also with a need for 13 bits at
>> base, and the D3X when in 12-bit mode. These are only
>> on the fringe of posterizing.
>>
>>> raising shadows in post for the deepest shadows, and
>>> in skies where the color pallet is very
>>> limited. Doesn't the noise level follow this same
>>> principle? Or is there an unrelated reason for the
>>> noise levels paralleling tone counts?
>> Any posterization you see in a RAW conversion is most
>> likely caused by the math used in the converter, and
>> nothing else. Of course, JPEG compression does some
>> posterization of its own, especially if you use too
>> much NR and it starts blocking up.
>
>OK, this makes sense, the posterizing issue is not really visible in any
>sort of normal exposure. What about Floyd's comment below that the noise
>level remains the same but exposing to the right increases the signal so
>that overwhelms the noise? That seems to tie the two together in a
>comprehensible way. The LL link just makes a lot of sense, it can't be
>complete BS. http://www.luminous-landscape.com/tutorials/expose-right.shtml

That is an *excellent* article. It also has a link to
another article, titled "Understanding Histograms",
which several contributors to this thread could benefit
from (particularly the histogram of the moon shot and
the high key tree image that are the last two shown at
the bottom of the article):

http://www.luminous-landscape.com/tutorials/understanding-series/understanding-histograms.shtml

--
Floyd L. Davidson <http://www.apaflo.com/floyd_davidson>
Ukpeagvik (Barrow, Alaska) floyd(a)apaflo.com
From: DRS on
"Floyd L. Davidson" <floyd(a)apaflo.com> wrote in message
news:87ab02ql1h.fld(a)apaflo.com
> "DRS" <drs(a)removethis.ihug.com.au> wrote:

[...]

>> I have always said it changes the indication of exposure. It is you
>> who made the false claim that I and others were saying it had
>> anything to do with changing the actual exposure, which we have
>> repeatedly denied.
>
> *All* *five* *show* *the* *exact* *same* *exposure*.

Yes, how many times do I have to tell you that before you get it?

> Not one of them shows any 1 f/stop difference as you
> claim.

Yes, they do, in terms of *accuracy*.

> The only thing that moves 1 f/stop are the tall peaks in
> the graph that have *nothing* to do with exposure. It
> is well to the left of the right edge, and it is that
> right edge that indicates exposure.

The right edge moves.

>> And because of the different contrast settings the 5 histograms show
>> the exposure differently. The exposure hasn't changed but the
>> accuracy of its representation has. Which is all that has been
>> claimed.
>
> They *don't* show the *exposure* differently.

Yes, they do. Everybody can see it except you.

>> None of the 5 histograms indicate overexposure. What they do
>> indicate, to different degrees of accuracy, is the room for exposure
>> compensation. Only 1 can be the most accurate and it is the -4
>> histogram.
>
> All of them indicate there is no room for any
> "compensation". Increasing the camera exposure will
> result in clipping, and each of those histograms shows
> that.

No, they don't. Try looking at what is there.

[...]

>> In this instance that indicates specular highlights, which as has
>> been noted by several people, may be blown without spoiling the
>> image. That is a choice by the photographer.
>
> The tall peaks are *not* a "specular highlights", and

I didn't say they were. Do you ever pay attention? You can't even follow
when I'm addressing your argument!

[...]

> You know, logically if what you said made sense the
> article would have discussed it in some way, but it does
> not even hint at it. If changing the contrast displayed
> a more accurate *exposure* (right edge location), why
> didn't they point to it?

"What does all this mean? In order to display the full dynamic range of
your camera in the histogram displayed on the back, you must set the
contrast to minimum. You now can see the same dynamic range in the
histogram on the camera as you will see it in your RAW processor on your
computer."



From: Porte Rouge on
On Oct 8, 6:22 am, fl...(a)apaflo.com (Floyd L. Davidson) wrote:
> Paul Furman <pa...@-edgehill.net> wrote:
> >John Sheehy wrote:
> >> Paul Furman wrote:
>
> >>> OK, the noise is more the reason but posterizing can
> >>> be a problem in dark areas...
> >> The only cameras I know of with even a hint of RAW
> >> posterization are the Pentax K10D, which would profit
> >> from 13 bits instead of 12 at ISO 100 (not for 200 or
> >> higher), the Sony A900 also with a need for 13 bits at
> >> base, and the D3X when in 12-bit mode.  These are only
> >> on the fringe of posterizing.
>
> >>> raising shadows in post for the deepest shadows, and
> >>> in skies where the color pallet is very
> >>> limited. Doesn't the noise level follow this same
> >>> principle? Or is there an unrelated reason for the
> >>> noise levels paralleling tone counts?
> >> Any posterization you see in a RAW conversion is most
> >> likely caused by the math used in the converter, and
> >> nothing else.  Of course, JPEG compression does some
> >> posterization of its own, especially if you use too
> >> much NR and it starts blocking up.
>
> >OK, this makes sense, the posterizing issue is not really visible in any
> >sort of normal exposure. What about Floyd's comment below that the noise
> >level remains the same but exposing to the right increases the signal so
> >that overwhelms the noise? That seems to tie the two together in a
> >comprehensible way. The LL link just makes a lot of sense, it can't be
> >complete BS.http://www.luminous-landscape.com/tutorials/expose-right.shtml
>
> That is an *excellent* article.  It also has a link to
> another article, titled "Understanding Histograms",
> which several contributors to this thread could benefit
> from (particularly the histogram of the moon shot and
> the high key tree image that are the last two shown at
> the bottom of the article):
>
>  http://www.luminous-landscape.com/tutorials/understanding-series/unde....
>
> --
> Floyd L. Davidson <http://www.apaflo.com/floyd_davidson>
> Ukpeagvik (Barrow, Alaska)              fl...(a)apaflo.com

Hey, that's the same article you said this about:

"But regardless of that, the cited URL above from
luminous-landscape is *not* full of good stuff. They
miss the point entirely, and provide nothing that is
actually useful!"