PSE ACR histogram highlight clipping point moves right with increased exposure

[digitaldog]

Here’s an image construct with four white patches:

http://digitaldog.net/files/AlmostWhiteGrad.tif

1st is RGB250 (because that value was mentioned), then 252, 253 and 254. NOT 255.

Open in ACR (you have to set preferences OR select “Format to Camera Raw”) so TIFF actually opens in ACR.

Turn ON the clipping indicator for highlights in Histogram. NO Red overlay. Because nothing is clipping to 255. That’s how the clipping indicator should and does work. Forget trying to assume what's the last so called 'clipping line' use a better tool to SEE if all three channels clip: the clipping overlay and/or the numbers!

Move Exposure slider to the right to read +0.5 or +0.10 or +0.15 or +0.20 or even +0.25. NO clipping from the RGB 254 square let alone 250!

Only until you move Exposure to +0.30 does that last value which was R254 move into clipping for all three channels which are now RGB 255 and the red overlay now comes onto the one,LAST square. And the RGB values are 255.

Nothing seen here is unexpected. 

The guess (“but it's not more than a guess“) that clipping is set to begin at around 250 is simply incorrect. Clipping is a well understood condition. For some. And RGB 250 isn't clipping. It certainly can after altering the RGB values as can many RGB values if pushed too far so it clips to RGB 255. RGB 255 clipping line is as far right of the Histogram as can be.

Side note: Raw CAN undergo highlight reconstruction (recovery) and how well is dependent on Process Version used. IF one channel of raw isn't clipped, but the other two are, ACR/LR can provide highlight recovery from the one channel. But NOT from anything done with Exposure slider being pushed to the right...

Enough said.

[Frans Waterlander]

I'm not sure I understand what the "Clipping line" is. Could you post a screenshot indicating what you call the Clipping line?

Here are the screenshots.

[digitaldog]

Here are the screenshots.

Of a single channel clipping in ACR.....

[Frans Waterlander]

Of a single channel clipping in ACR.....

And what, may I asked, is the significance of that remark?

[digitaldog]

And what, may I asked, is the significance of that remark?

That two are not and this is an attribute of saturation clipping of one channel.
And yet, you started this 'discussion' writing:
"When you increase exposure in the Photoshop Elements ACR converter, the highlight clipping point shows up when you start to blow out the highlights, as you would expect".
Do you understand the differences between clipping highlights and clipping a single channel?
Really, the way ACR treats highlight clipping was shown with a test image and it works as expected.
And the push back begins Jeremy.
I give up. Proof of concept was provided. I can do no more (nor should I have to).

[Frans Waterlander]

Do you understand the differences between clipping highlights and clipping a single channel?

Do you? Take a look at the ACR histogram when the Exposure slider is increased by one step, to -1.05. Still only one channel is clipping, but now the clipping warning is turned on. Hmmmm... Comments?

[digitaldog]

Do you? Take a look at the ACR histogram when the Exposure slider is increased by one step, to -1.05. Still only one channel is clipping, but now the clipping warning is turned on. Hmmmm... Comments?

Yes, you've confirmed you don't understand what the red ("Clipping line" your invented term) indicates and that the Exposure slider in this setting isn't (yet) clipping the highlights (all channels) and that you don't know what that warning is telling you and that you didn't watch a video that would have taught you how this all works, that you didn't download a test image that shows how ACR works as those of us who know, works as expected, and that your guess about clipping at 250 is just that, a guess and and wrong, that you are again wasting my time with your push back. Man, you've told us a lot about where you're heading in this thread. As I predicted in the 2nd post here.
Hmmmm indeed. 
“There are two ways to be fooled. One is to believe what isn't true; the other is to refuse to believe what is true.” -Søren Kierkegaard

[digitaldog]

Take a look at the ACR histogram when the Exposure slider is increased by one step, to -1.05. Still only one channel is clipping, but now the clipping warning is turned on. Hmmmm...

Take a look at the ACR histogram when YOUR Exposure slider is DECREASED by one STOP, to -1.05.   

Enough said. Again. For Jeremy's sake and so you are not further embarrassed by facts, and because again you are asking questions where the predetermined answers have to fit your acceptance instead of the facts, I bow out and will turn off notifications of this thread. Maybe someone will join in, (those two other members?) or otherwise and attempt to explain how this works, either factually and with your pushbacks, as I have, or as a fabrication to fit your desires.
“Facts do not cease to exist because they are ignored.”
― Aldous Huxley

[Frans Waterlander]

Yes, I may have invented the term "clipping line", so sue me. Fact is, at the Exposure slider value of -1.75, there is no such line, no pixels stacking up. But increase Exposure beyond -1.75 and the line appears at a guesstimated value of 250 and grows in the same position until you reach Exposure value -1.30 and then the line starts to move until it reaches the maximum value at which point the clipping warning turns on. Nothing of what you have said sofar explains this weird behavior.

[fdisilvestro]

Hi Frans,

Thanks for posting the screenshots. What is happening here is a different issue. Your raw image has a channel severely clipped and what you see in the histogram is not a "clipping line" but a bunch of pixels at the same value.

A clipped channel is bad because there is no way to know how clipped it is, so any highlight reconstruction algorithm has to "guess" what to do with those clipped values, and the LR/ACR algorithm does what you are seeing in the histogram. For instance, the software RawTherapee has different highlight reconstruction algorithms to choose from, with different outcomes depending on the case.

As you continue to move the exposure slider to the left, the quantity of pixels at the same level with decrease, due to the math involved, including, but not limited to rounding issues. This is what you are referring as if the line disappear when you have exposure of -1.75

You could check the raw file in a serious histogram tool like RawDigger to see the clipped channel.

[Frans Waterlander]

Hi Franciso,

I understand all that and yes, I know that the red channel gets severely clipped when I increase the Exposure slider setting beyond the point where it start to clip. My point is that the line showing the clipped red channel is not stationary but moves to the right as the Exposure increases and that doesn't make any sense to me.

[fdisilvestro]

Hi Franciso,

I understand all that. My point is that the line showing the clipped red channel is not stationary but moves to the right as the Exposure increases and that doesn't make any sense to me.

That's the way the highlight reconstruction algorithm works. In ACR it gives priority to keeping the tone & neutrals. It is not a clipping line, it is just a large amount of pixels with the same value or inside a very narrow range of values. In this case you are not increasing exposure, but reducing negative exposure (it seems a small difference but it is not!), so all that bunch of pixels in a narrow range will move to the right until they reach 255

Use a raw image that does not have clipped values and you will see that you cannot replicate this behaviour.

[Frans Waterlander]

That's the way the highlight reconstruction algorithm works. In ACR it gives priority to keeping the tone & neutrals. It is not a clipping line, it is just a large amount of pixels with the same value or inside a very narrow range of values. In this case you are not increasing exposure, but reducing negative exposure (it seems a small difference but it is not!), so all that bunch of pixels in a narrow range will move to the right until they reach 255

Use a raw image that does not have clipped values and you will see that you cannot replicate this behaviour.

That makes a lot of sense, Francisco. Thanks a bunch! And yes, I don't see this behavior with "normal" images.

[bjanes]

Right on! This is not an issue of whether or not the histogram in the Adobe Camera RAW (ACR) converter faithfully represents the underlying data, but why the highlight clipping level is shifted to the right when the Exposure slider is moved to the right (as confirmed by using Levels after conversion).

The current Process Version 5 in ACR and Lightroom is a poor tool for evaluating highlight clipping in the raw file as it invokes automatic highlight recovery. To get a better indication of the highlights you can use Rawdigger, which gives an accurate raw histogram. In ACR and LR you should repeat your experiment using Process Ver 2 with the sliders on the main panel set to zero and the point curve set to linear. Adobe applies a baseline exposure offset for exposure (see here) for details. For the Nikon D850 the offset is +0.35 EV, so you should set the exposure to -0.35 EV to compensate for the offset. With these settings ACR/LR gives a more accurate indication of the raw file values.

As you increase the exposure in PV 5 automatic highlight recovery corrects for highlight clipping up to a certain extent until it can no longer do so and the histogram shows clipping. Compare this clipping point with that shown by PV 2 which shows clipping considerably earlier and the behavior of the histogram may be more like what you expected. In any case, I do not see why you should make such a fuss over this issue. I don't see any significant anomaly. You should post images showing the behavior you allege.

Bill

[Frans Waterlander]

Hi Bill,

Thanks for your inputs. I'll play around with different process versions. I raised the issue because the ACR histogram at various Exposure values didn't make sense to me. Thanks to you and Francisco it now does.
My previous posts have screenshot that show the behavior I described.

[Guillermo Luijk]

Do you? Take a look at the ACR histogram when the Exposure slider is increased by one step, to -1.05. Still only one channel is clipping, but now the clipping warning is turned on. Hmmmm... Comments?

Frans, you seem to be concerned about highlight clipping, but I think the problem here is not clipping because of exposure but because out of gamut colour conversion. Look at your shadows: blue channel clipped, and highlights: red chanel clipped. Your image has some saturation that your current output colour profile chosen (was it sRGB? try to switch to ProPhoto RGB and look at the histogram) cannot handle.

Regarding real highlight clipping, it's very rare to have the red channel clipped and the green channel intact. Almost always on any camera and scene, the green channel is the first to clip on the RAW file.

If you can upload that RAW file we can find out. I'd bet ACR is just showing you an out of gamut issue, and what he (or it? or she?) is telling you is correct.

Regards

[digitaldog]

I'd bet ACR is just showing you an out of gamut issue, and what he (or it? or she?) is telling you is correct

Indeed, he was told this (« Reply #25) and promptly ignored it. As will our posts. 

[Frans Waterlander]

Frans, you seem to be concerned about highlight clipping, but I think the problem here is not clipping because of exposure but because out of gamut colour conversion. Look at your shadows: blue channel clipped, and highlights: red chanel clipped. Your image has some saturation that your current output colour profile chosen (was it sRGB? try to switch to ProPhoto RGB and look at the histogram) cannot handle.

Regarding real highlight clipping, it's very rare to have the red channel clipped and the green channel intact. Almost always on any camera and scene, the green channel is the first to clip on the RAW file.

If you can upload that RAW file we can find out. I'd bet ACR is just showing you an out of gamut issue, and what he (or it? or she?) is telling you is correct.

Regards

Hi Guillermo,

I wasn't concerned about highlight clipping by or in itself, but rather about the looks of the ACR histogram. As others have explained, what you see in this particular case is not a varying clipping value, but highlight recovery at work. Once you know that, all makes sense.

The histogram looks non-traditional because the image is non-traditional: it's an infrared image, so oodles of values in the red channel, less so in the green and way, way less in the blue.

[Frans Waterlander]

Indeed, he was told this (« Reply #25) and promptly ignored it. As will our posts. 

Reply #25 on my screen is one of my own replies. I seem to remember we went through this before where your reply numbers didn't match with mine. Could you say what day and time for the reply you are referring to?

[bjanes]

Frans, you seem to be concerned about highlight clipping, but I think the problem here is not clipping because of exposure but because out of gamut colour conversion. Look at your shadows: blue channel clipped, and highlights: red chanel clipped. Your image has some saturation that your current output colour profile chosen (was it sRGB? try to switch to ProPhoto RGB and look at the histogram) cannot handle.

Regarding real highlight clipping, it's very rare to have the red channel clipped and the green channel intact. Almost always on any camera and scene, the green channel is the first to clip on the RAW file.

If you can upload that RAW file we can find out. I'd bet ACR is just showing you an out of gamut issue, and what he (or it? or she?) is telling you is correct.

Regards

+1

Without the proper context, histograms can be quite misleading. Here is an example. A red flower was photographed with the Nikon D850 with the camera set to AdobeRGB and the Picture Control to flat.

The camera histogram shows a nearly ideal ETTR exposure. Looking at the histogram in ACR Process Version 5 with the Adobe Color profile and rendering into AdobeRGB, the histogram is slightly to the left of the camera histogram. However, looking at the raw histogram in Rawdigger demonstrates that the red channel is more than 2 EV short of clipping. The white balance coefficients are added to the Rawdigger histogram. Note that the red channel is multiplied by 1.90 for white balance.

Finally, rendering into sRGB with the same settings in ACR shows prominent clipping in the red channel.

Bill