Ad
Ad
Ad
Pages: [1]   Bottom of Page
Print
Author Topic: ACR histogram confusion?  (Read 12312 times)
Redcrown
Full Member
***
Offline Offline

Posts: 143


« on: September 27, 2013, 12:43:59 PM »
ReplyReply

Playing with an image that has some difficult bright red tones, I noticed something in ACR that confuses me. The ACR histogram of the image shows clipping of the red channel when, in fact, there are no red channel values above about 235.

But this happens only when ACR is set to Adobe98 or sRGB colorspace. If ACR is set to ProPhoto, the histogram shows no clipping of the red channel.

I understand that the image probably has non-clipped red values that are outside of the Adobe98/sRGB colorspaces. And I understand that those values will be brought inside the output colorspace during the conversion. But I don't understand why ACR is showing a clipped histogram. It appears that the ACR histogram is actually showing out-of-gamut values as opposed to true RGB values.

What am I missing here? I always thought a histogram shows where values fall across the 0 to 255 range, regardless of where they fall within a particular gamut.

Here is a composite screen cap of the image I'm using. It's actually a small cropped portion of a bright red wig.


Logged
Tony Jay
Sr. Member
****
Offline Offline

Posts: 2147


« Reply #1 on: September 27, 2013, 04:01:31 PM »
ReplyReply


...What am I missing here? I always thought a histogram shows where values fall across the 0 to 255 range, regardless of where they fall within a particular gamut.



Huh!
Unless the histogram reflects the reality of the colourspace assigned to the image then the information is meaningless.
It is exactly appropriate that those reds so comfortably accommodated in ProPhotoRGB might clip if the image is assigned the Adobe98 colourspace.

There is no "ABSOLUTE" colour reference to which all colourspaces are linked.
For practical purposes ProPhotoRGB is used in this way because it is the biggest easily accessible colourspace.
AdobeRGB and sRGB fit comfortably inside ProPhotoRGB.

Tony Jay
Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 9192



WWW
« Reply #2 on: September 27, 2013, 04:32:21 PM »
ReplyReply

But this happens only when ACR is set to Adobe98 or sRGB colorspace. If ACR is set to ProPhoto, the histogram shows no clipping of the red channel.
That is an out of gamut clipping. With the current rendering in Adobe RGB (1998) or sRGB, that red is OOG. But as soon as you select ProPhoto RGB, that color space is large enough that you are not clipping the gamut. This is a good illustration why ProPhoto RGB is a better working space for raw workflows, at least when you hand it an image who's scene gamut is large.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
Redcrown
Full Member
***
Offline Offline

Posts: 143


« Reply #3 on: September 27, 2013, 11:22:35 PM »
ReplyReply

Thanks guys, but let me come at this another way.

When I have this test image in ACR and set to Adobe98 colorspace I see the clipped histogram as shown above. I've always assumed that that red spike on the right side of the histogram represented pixels whose red value is clipped at 255.

So I grab the color sampler and fish around on the image looking for those clipped pixels. But I can't find them. I can't find any place where the red value is greater than 235.

What does that mean? Are the RGB readouts of the color sampler showing values AFTER conversion to Adobe98 (after values have been mapped into the target colorspace), while the histogram is showing the original ProPhoto RGB values assigned to Adobe98?

Here is another point of curiosity. If I convert the raw as 16bit Prophoto, save that as a TIF, then open that TIF in ACR, I get a significantly different histogram. The histogram I see in ACR is not what I get in Photoshop. So what, exactly, is the ACR histogam showing me?


Logged
Tony Jay
Sr. Member
****
Offline Offline

Posts: 2147


« Reply #4 on: September 28, 2013, 03:51:18 AM »
ReplyReply

The scale 0-255 is not an absolute scale with defined "distances" in a histogram.
So, ProPhotoRGB Adobe98 RGB and sRGB will all show a histogram with values between 0-255 but the RGB values derived from a sRGB histogram will not be the same as that derived from an AdobeRGB or ProPhotoRGB histogram even though the actual colour is the same.
So those values are only relative to the gamut from which they are derived.
Component RGB colours outside the gamut of the colourspace in question will be assigned 0 or 255.
So a red too saturated for a particular colourspace will have a red value of 255.
(In fact shades of red that can be distinguished with the naked eye can all read 255 as long as the monitor on which they are being viewed is capable of displaying them as different - and hopefully accurate - colours and the gamut is small enough.)
In another colourspace that can accomodate that particular shade of red the red value may read as 223.
That is why a ProphotoRGB histogram will graphically show a red as being not saturated and mousing over that shade of red shows a value <255 while the same red shade in a sRGB histogram shows graphically as being saturated and mousing over the same area on the image will show a red value of 255.

I think a trip through resources dealing with colourspaces may be in order.

Tony Jay
« Last Edit: September 28, 2013, 04:07:37 AM by Tony Jay » Logged
bjanes
Sr. Member
****
Online Online

Posts: 2838



« Reply #5 on: September 28, 2013, 11:07:09 AM »
ReplyReply

The scale 0-255 is not an absolute scale with defined "distances" in a histogram.
So, ProPhotoRGB Adobe98 RGB and sRGB will all show a histogram with values between 0-255 but the RGB values derived from a sRGB histogram will not be the same as that derived from an AdobeRGB or ProPhotoRGB histogram even though the actual colour is the same.
So those values are only relative to the gamut from which they are derived.

To expand on what Tony has said, the color values in  Adobe RGB and ProPhotoRGB are not absolute, but are relative to the space in which they are defined. The L*a*b space is device independent. However, when one renders into L*a*b, the values are not what was present in the actual scene but depends on how they were rendered into the space.

When evaluating the histograms for clipping, it is helpful to determine if clipping was present in the raw file or occurred via overzealous development settings. The first step is to look at the histogram of the raw file. Rawdigger is an excellent tool for this purpose. Clipping in the rendered file can be due to luminance or saturation clipping. Generally speaking, saturation clipping in ACR is indicated by one or more of the color channels showing a cliff like cutoff on the right of the histogram.

Here is a scene rendered into ProPhotoRGB with ACR. The red channel is short of clipping.


The raw histogram shows that all channels are exposed to the right without any clipping. The red channel is considerably short of clipping, but when white balance is applied, the red histogram will move to the right.


Rendering into Adobe RGB with the same settings shows saturation clipping of the red channel.


With the latest version of ACR, one can render into L*a*b, which separates luminance and color and is device independent. The histogram is here and shows a full range of luminance without clipping. The color channels are far from clipping.


Overexposure rendering into ProPhotoRGB does show clipping, which is largely luminance clipping. The red channel appears minimally clipped.


The L*a*b histogram shows that the color channels are free of clipping, while the luminance is clipped.


Bill
Logged
mouse
Full Member
***
Offline Offline

Posts: 171


« Reply #6 on: September 28, 2013, 04:56:18 PM »
ReplyReply

I have read the replies to Redcrown's post, and they are all very clear and easily understandable.
But, one of his questions remains unanswered:

Quote
When I have this test image in ACR and set to Adobe98 colorspace I see the clipped histogram as shown above. I've always assumed that that red spike on the right side of the histogram represented pixels whose red value is clipped at 255.

So I grab the color sampler and fish around on the image looking for those clipped pixels. But I can't find them. I can't find any place where the red value is greater than 235.

If I correctly understand Tony's reply:

Quote
Component RGB colours outside the gamut of the colourspace in question will be assigned 0 or 255.

OOG colors will appear as a spike in the histogram (values = 255) even though their luminance values may be <255.  But shouldn't these same pixels show up with values = 255 when searched for with the color sampler?  Is it simply because the histogram reflects the "gamut status" of these pixels, but the color sampler responds only to luminance?
Logged
bjanes
Sr. Member
****
Online Online

Posts: 2838



« Reply #7 on: September 29, 2013, 07:19:14 AM »
ReplyReply

I have read the replies to Redcrown's post, and they are all very clear and easily understandable.
But, one of his questions remains unanswered:

If I correctly understand Tony's reply:

OOG colors will appear as a spike in the histogram (values = 255) even though their luminance values may be <255.  But shouldn't these same pixels show up with values = 255 when searched for with the color sampler?  Is it simply because the histogram reflects the "gamut status" of these pixels, but the color sampler responds only to luminance?

An interesting thought, but ACR does not have luminosity histograms or luminosity readouts. The histograms and readouts are only RGB. For an explanation of luminosity histograms, interested readers should look at the histogram tutorial in Cambridge in Color. A luminosity of 255 requires the presence of white (255,255,255). Photoshop does have a luminosity histogram, but the info tool gives only RGB values and does not show luminosity.

If the rendered image shows channel clipping, the clipped channels must have values at 255.

Regards,

Bill
Logged
Vladimirovich
Sr. Member
****
Offline Offline

Posts: 1320


« Reply #8 on: September 29, 2013, 08:50:33 AM »
ReplyReply

An interesting thought, but ACR does not have luminosity histograms or luminosity readouts. The histograms and readouts are only RGB.
ACR has now Lab and CMYK histogram/readouts (you can select a color space for ACR output now), but to call what ACR displays a histogram will be an insult... ungraduated, etc.
Logged
bjanes
Sr. Member
****
Online Online

Posts: 2838



« Reply #9 on: September 29, 2013, 10:07:34 AM »
ReplyReply

ACR has now Lab and CMYK histogram/readouts (you can select a color space for ACR output now), but to call what ACR displays a histogram will be an insult... ungraduated, etc.

Vladimirovich,

You seem to be easily insulted. The Rawdigger histograms are exemplary since they identify the scale for both the ordinate and abscissa and give the relevant statistics. I prefer the abscissa to have a log2 scale to show EV values.  The ACR histograms do not supply this information and have no scaling tics, but they serve their purpose for evaluating an image. The abscissa appears to be scaled with gamma encoded values.



Regards,

Bill
Logged
Vladimirovich
Sr. Member
****
Offline Offline

Posts: 1320


« Reply #10 on: September 29, 2013, 11:11:17 AM »
ReplyReply

Vladimirovich,

You seem to be easily insulted.

that was a "figure of speech"...
Logged
mouse
Full Member
***
Offline Offline

Posts: 171


« Reply #11 on: October 03, 2013, 02:39:33 PM »
ReplyReply

ACR has now Lab and CMYK histogram/readouts (you can select a color space for ACR output now),

I am running ACR v8.2 (PSCS6).  I cannot select Lab or CMYK color space for ACR output.  Options are ProPhoto, Adobe RGB, sRGB and Color Match. 

Is this ability found only in the Cloud?
Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 9192



WWW
« Reply #12 on: October 03, 2013, 02:42:42 PM »
ReplyReply

Options are ProPhoto, Adobe RGB, sRGB and Color Match. 

In Space dropdown menu? You should see multiple color spaces including Lab.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
mouse
Full Member
***
Offline Offline

Posts: 171


« Reply #13 on: October 03, 2013, 08:21:22 PM »
ReplyReply

In Space dropdown menu? You should see multiple color spaces including Lab.

Based on Vladomirovich's post, that is what I expected.  But the dropdown menu shows only the 4 spaces I listed.  Where am I going wrong.  Is it possible that one needs the latest Cloud version to obtain this option?  I am running version 8.2.
Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 9192



WWW
« Reply #14 on: October 03, 2013, 08:36:34 PM »
ReplyReply

Based on Vladomirovich's post, that is what I expected.  But the dropdown menu shows only the 4 spaces I listed.  Where am I going wrong.  Is it possible that one needs the latest Cloud version to obtain this option?  I am running version 8.2.

You should see this (so be sure you're actually launching ACR 8.2):
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
Vladimirovich
Sr. Member
****
Offline Offline

Posts: 1320


« Reply #15 on: October 04, 2013, 08:39:38 AM »
ReplyReply

You should see this (so be sure you're actually launching ACR 8.2):
and ACR shall be launched by "CC" PS or Bridge
Logged
Vladimirovich
Sr. Member
****
Offline Offline

Posts: 1320


« Reply #16 on: October 04, 2013, 08:40:45 AM »
ReplyReply

I am running version 8.2.
you need "CC" versions of PS or Bridge, otherwise new functionality is blocked
Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 9192



WWW
« Reply #17 on: October 04, 2013, 08:44:16 AM »
ReplyReply

you need "CC" versions of PS or Bridge, otherwise new functionality is blocked
Oh absolutely! Missed that he was using CS6.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
mouse
Full Member
***
Offline Offline

Posts: 171


« Reply #18 on: October 04, 2013, 02:19:03 PM »
ReplyReply

Thanks for the clarification.
I am NOT in the Cloud, so that explains it.
Logged
kers
Sr. Member
****
Offline Offline

Posts: 773


WWW
« Reply #19 on: October 04, 2013, 05:15:30 PM »
ReplyReply

Thanks for the clarification.
I am NOT in the Cloud, so that explains it.

sounds better ; to be not in the cloud ( with your head) ...Smiley

still the LAB option is nice
Logged

Pieter Kers
www.beeld.nu
Pages: [1]   Top of Page
Print
Jump to:  

Ad
Ad
Ad