Pages: 1 [2]   Go Down

Author Topic: What Does This Histogram Mean?  (Read 21502 times)

JohnKoerner

  • Guest
Re: What Does This Histogram Mean?
« Reply #20 on: August 22, 2011, 05:52:27 am »

Okay, I undertstand now, thank you.

Jack


.
Logged

Ellis Vener

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2151
    • http://www.ellisvener.com
Re: What Does This Histogram Mean?
« Reply #21 on: August 22, 2011, 09:23:31 am »




Does this mean I have no shadow clipping, no highlight clipping, but DO have color clipping that is outside the Adobe RGB color space?

Forgive the simplicity of the question, but I am trying to better understand what the histogram is telling me.

Would this be an instance where, say, a "print" rendered in the ProPhoto color space would give me a substantially more "vivid" output, color-wise, than what I am seeing on my Adobe RGB screen?

Thanks for any help,

Jack

.
To break it down using simple language:

"Does this mean I have no shadow clipping, no highlight clipping"
Yes.

"but DO have color clipping that is outside the Adobe RGB color space?"
The Lightroom histogram doesn't tell you anything about how this  image will render when converted to Adobe RGB (1998), any other color space (including ProPhoto RGB) or in any device (display or printing) profile.

"I am trying to better understand what the histogram is telling me."  It does however give a pretty good idea of how the tonal values are dispersed in your raw image. A histogram is a bar graph

"Would this be an instance where, say, a "print" rendered in the ProPhoto color space would give me a substantially more "vivid" output, color-wise"

As far as I have been able to find out in my readings and experiments with various printer, ink and paper combinations there are no printing processes  currently capable of rendering , even to the limits of human vision*, the full gamut of ProPhoto RGB (* being a synthetic mathematically calculated ideal , the gamut of ProPhoto RGB extends beyond the gamut of even the healthiest human eye. ). There are however certain printer, ink and paper (which jargoneers will call substrates) which have gamuts that in some very saturated mid- tone colors are larger than what Adobe RGB(1998) is capable of containing.

"than what I am seeing on my Adobe RGB screen"
Like ProPhoto RGB and sRGB, Adobe RGB(1998) is a device independent (some will say relatively device independent) synthetic colro space. Your display isa real world device. It may be capable of showing all colors in Adobe RGB(1998) but it really needs its own profile hopefully made wit ha very high quality colorimeter or photospectrometer to determine the way the mechanics and electronics of your specific individual video card and display  combination render individual colors and their relationships with each other.

There seem to be many here in this forum especially who like to argue not only how many angels can dance on the head of a pin but what shape those angels take, but in the end as people like John Paul Caponigro. MAc holbert, BillAtkinson, and any good printer worth his salt will point out: the print is the real proof.

"Thanks for any help,"

Hopefully my post is actually of practical help to you.  A couple of years ago I wrote a short article about histograms: http://www.ppmag.com/web-exclusives/2007/12/what-is-a-histogram-and-how-do.html
« Last Edit: August 22, 2011, 09:30:14 am by Ellis Vener »
Logged

MarkM

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 428
    • Alaska Photographer Mark Meyer
Re: What Does This Histogram Mean?
« Reply #22 on: August 22, 2011, 02:28:51 pm »

The Lightroom histogram doesn't tell you anything about how this  image will render when converted to Adobe RGB (1998), any other color space (including ProPhoto RGB)…

Ellis, I wonder if you would elaborate on this. It's my understanding that the histogram in lightroom is based on ProPhotoRGB (or the Melissa RGB variant) and the histogram would in fact tell you quite a bit about the image in the prophoto space.
Logged

Ellis Vener

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2151
    • http://www.ellisvener.com
Re: What Does This Histogram Mean?
« Reply #23 on: August 22, 2011, 02:52:18 pm »

Ellis, I wonder if you would elaborate on this. It's my understanding that the histogram in lightroom is based on ProPhotoRGB (or the Melissa RGB variant) and the histogram would in fact tell you quite a bit about the image in the prophoto space.
You are correct: it  does tells you a lot about tonal distribution in ProPhoto RGB , but just not the whole story. For starters "Melissa RGB" (Lightroom's native color space ) is based on ProPhoto RGB but ProPhoto RGB has a gamma of 1.8 while Melissa's uses a more perceptually uniform 2.2 gamma.  I believe ( I don't have my notes i nfront of me) that Melissa RGB's tonal response is modeled on sRGB's.
« Last Edit: August 23, 2011, 11:34:10 pm by Ellis Vener »
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: What Does This Histogram Mean?
« Reply #24 on: August 22, 2011, 03:01:10 pm »

While MelissaRGB and ProPhoto are close due to the same primaries, the question becomes, what’s the difference on a Histogram due to the gamma encoding.

Its pretty easy to test this in Photoshop. First you need to make a custom RGB working space of MelissaRGB. Just select ProPhoto and then go to the Custom menu in the Color Settings, enter the 2.2 gamma value and save it. Then you can convert into both. In the example here, I have a Granger Rainbow built in Lab and converted to both. Close but not the same. How the differences are or are not important is up to debate:

Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

MarkM

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 428
    • Alaska Photographer Mark Meyer
Re: What Does This Histogram Mean?
« Reply #25 on: August 22, 2011, 03:31:03 pm »

Yes, I understand that much, but the gamma value should have no affect on the endpoints of the histogram. Since lightroom shows individual channels, it will tell us about tonal distribution, obviously, but shouldn't it also inform us about clipping colors that exceed the gamut of prophoto rgb?

If you import an image into lightroom with out of gamut colors, you should see this in the histogram as individual channels getting clipped. For instance the attached single color chip which is outside ProPhoto RGB.
Logged

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: What Does This Histogram Mean?
« Reply #26 on: August 22, 2011, 04:28:32 pm »

Yo uare correct: it  does tells you a lot about tonal distribution in ProPhoto RGB , but just not the whole story. For starters "Melissa RGB" (Lightroom's native color space ) is based on ProPhoto RGB but ProPhoto RGB has a gamma of 1.8 while Melissa's uses a more perceptually uniform 2.2 gamma.  I believe ( I don't have my notes i nfront of me) that Melissa RGB's tonal response is modeled on sRGB's.
Melissa is somewhat confusiong in Lightroom. The working space is linear (gamma = 1), but the histograms and info readouts are in terms of an sRBG tone curve, which is gamma 2.2 with a linear segment for the deep shadows. The tonal distribution for a linear space is suboptimal, since the quantization steps are too large for shadow values (see encoding, by Greg Ward, Figure 2 and text). Using a gamma of 2.2 improves the quantization in the shadows, but a linear segment is still needed for the deep shadows as the quantization step approaches infinity as the luminance approaches zero.

Regards,

Bill
Logged

Ellis Vener

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2151
    • http://www.ellisvener.com
Re: What Does This Histogram Mean?
« Reply #27 on: August 23, 2011, 08:34:29 am »

Melissa is somewhat confusiong in Lightroom. The working space is linear (gamma = 1), but the histograms and info readouts are in terms of an sRBG tone curve, which is gamma 2.2 with a linear segment for the deep shadows. The tonal distribution for a linear space is suboptimal, since the quantization steps are too large for shadow values (see encoding, by Greg Ward, Figure 2 and text). Using a gamma of 2.2 improves the quantization in the shadows, but a linear segment is still needed for the deep shadows as the quantization step approaches infinity as the luminance approaches zero.

Regards,

Bill
Thanks for that explanation Bill.

Ellis
Logged

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2005
    • http://www.guillermoluijk.com
Re: What Does This Histogram Mean?
« Reply #28 on: September 02, 2011, 07:11:21 pm »

The tonal distribution for a linear space is suboptimal

That's true only for an integer encoding. Floating point formats fix this, since they devote a number of bits to the significant digits and some bits to the exponent, so representation of deep shadows is as good as highlights.

Effect of gamma on the histogram:

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: What Does This Histogram Mean?
« Reply #29 on: September 03, 2011, 07:17:46 am »

That's true only for an integer encoding. Floating point formats fix this, since they devote a number of bits to the significant digits and some bits to the exponent, so representation of deep shadows is as good as highlights.

Yes, that is true, and thanks for posting the histograms showing how data are crowded at the low end with linear integer encoding and how the situation is improved with gamma encoding. However, IEEE 32 bit floats are overkill for digital imaging and require 96 bits/pixel. Photoshop uses 32 bit floats for HDR, presumably because that is what the computer hardware supports. Log encoding has constant error and floating point nearly so, as shown in Figure 3 of Greg Ward's excellent paper on encoding. Current ADCs used for photography have linear integer encoding, but log ADCs are widely used in communications. However, linear integer encoding is sufficient for current cameras, since gaps in the shadows are dithered by noise and are not apparent in images.

However, 16 bit integer encoding is insufficient for true High Definition (HDR) and Greg discusses various log and floating point HDR encodings in his article. The Digital Light and Color OpenEXR uses half precision floats (16 bits--48 bits/pixel) for HDR. As our sensors improve and get better dynamic range, it will probably be necessary to use some type of HDR encoding.

Regards,

Bill
Logged

Waeshael

  • Newbie
  • *
  • Offline Offline
  • Posts: 20
  • Medical camera designer since 1976
    • God, dogs and photography
Re: What Does This Histogram Mean?
« Reply #30 on: September 11, 2011, 10:28:07 pm »

Where did the picture come from? What was the color space assigned to the picture in the editor? If the capture was in "camera color space" and it was converted to sRGB choosing either relative colorimetric or perceptual, then all the colors in the scene will be moved into the color space selected. There will be no clipping of colors. The colors may not represent the scene exactly but the colors will be represented within the color space. Only if absolute colorimetric is chosen will the colors that are outside the chosen color space be clipped.
If you want to maintain the closest relationship between the scene and the image using absolute colorimetric, then you must convert the camera data from RAW into a color space big enough to hold all the colors. For many people this is Adobe 1998. But that is only done for product shots (absolute colorimetric, I mean.)

The histogram only shows colors that are available to it - which is the monitor color space usually close to sRGB - so no matter how much you stare at it you aren't going to see any colors beyond the small sRGB color space of your monitor, which hopefully is calibrated. If you start with Adobe 1998 color space, the CMS converts it into monitor color space by the conversion method you have chosen (perceptual, relative colorimetric or absolute colorimetric.) If you don't choose absolute colorimetric then no color data is clipped, only changed to fit what the monitor is able to show - that is good greens and blues and not so good reds and yellows.
You have to learn from printing pictures what the colors on the monitor will look like when they are printed - look at a dried print (after 24 hrs.) Nothing in the histogram is going to show you what the final print colors will look like.
So what can you tell from the histogram?

Remember all you see is a representation of the colors that the monitor can produce, not what the printer can print. IF you have an image converted to sRGB from camera RGB, and if you see that the left and right edges of the histogram bump up against the edges of the graph, then you know that sRGB colorspace is being used to the max. That is the darkest and lightest tones of each color are exactly fitting the color space you have chosen. This sounds good eh?
But what if in your editing you decide to boost the color a little, or to lighten the image? WHat happens is that perhaps one of the color channels get stopped from increasing, and yet the others can increase. Green, for instance, might become a little more blue than it was when is was darker.
So what to do?
Instead of converting to sRGB from camera RGB, try converting to Adobe 1998. Now what you see is that the histogram takes up less space on the graph, and there is "breathing room" at each end. This is the room you need in order to make changes to the image without clipping any of the color channels. So this new histogram shows the same SRGB data as before but in a new bigger "envelope." This is the situation you showed for the spider picture. But you still cannot see any colors other than sRGB - you still can't see the deep blues and reds, and bright yellows because though the printer can print them the monitor can't display them. This is due to of a limitation of the LED matrix of RGBG diodes. It's hard to make yellows out of red blue and green diodes - yellow comes out a shade of orange. Reds are okay as long as they are medium red to bright red, but dark red looks like mud on a monitor.

 If you want to emphasize the yellows, say, brighten up the yellow in the image editor knowing that the printer will respond to the changes more than the monitor. Check your histogram to make sure that it hasn't reached the edge of the graph after the adjustment. If it has; back off a little with the yellow emphasis. In your editor, convert your image color profile to that of the ink and paper, and send this direct to the printer with software color management turned off, and no adjustments of the printer colors - just let the printer handle the color management. Now look at the print after it has dried (24 hrs) and see if it is what you wanted. If not tweak the yellow again. Repeat until you get what you want.

You see, the printer will pull the colors into its own color space for that ink and paper combination. It won't clip colors.

Now some people choose to soft proof the image and try to see whether colors will be clipped on printing, but this test is done using the computer CMS, not the printers, and we aren't going to use the computer CMS are we?

So the only way to see what will happen to your lovely spider is to print, print, and print again.

Great picture!

By the way: in nature most colors are rather muted and easily fit within the sRGB space. Color films like Velvia and Kodachrome oversaturated the colors so that they printed nice in travel magazines. These films have a wide gamut which exceeds sRGB by a long way. That's why Adobe 1998, and Ektaspace PS 5.0 were developed - to hold all the oversaturated colors captured on film. Scanners were built with wide gamuts (wider than Adobe 1998) and files saved into big color spaces. Digital cameras don't have to deal with such a wide range of color - except in product shots in the studio, so sRGB is usually okay as a capture color space - but the working color space should be bigger - convert into it as a first step.

So my advice is to make your decisions based on what you see in the picture. Adjust your pictures by eye - not by numbers. Print until it looks good, adjusting colors by eye. If you were printing big pictures costing hundreds of dollars in paper and ink, that is a different story - go to the pro printers like Joe Holmes or get advice from Gerald Bybee who though "retired" from printing to a vineyard he is still shooting and knows all about printing.

cheers

Waeshael.
Logged
Digital technology is fun.

MarkM

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 428
    • Alaska Photographer Mark Meyer
Re: What Does This Histogram Mean?
« Reply #31 on: September 12, 2011, 12:35:54 am »

The histogram only shows colors that are available to it - which is the monitor color space usually close to sRGB

Kind of a lot to digest, but it's hard to get past this, which is just not true. The histogram shows data relative to the working space, not the monitor color space. I have two monitors of vastly different capabilities attached to this work station. I assure you the histogram does not change when I move from one monitor to the other.
Logged

Ellis Vener

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2151
    • http://www.ellisvener.com
Re: What Does This Histogram Mean?
« Reply #32 on: September 12, 2011, 08:47:01 am »

I see in your signature that you have been designing medical cameras since 1976 so I'd like to better understand where you are coming from with your advice as most of us don't design cameras.

I have questions about some aspects of your posts as from my understanding of things you are starting from false premises.

 "If the capture was in "camera color space" and it was converted to sRGB choosing either relative colorimetric or perceptual, then all the colors in the scene will be moved into the color space selected. There will be no clipping of colors. The colors may not represent the scene exactly ... "

The colors in a scene exist independently from the limitations of the sensor, processor and processing algorithms  used in a camera, and also from the artificial constraints of any RGB, LAB, XYZ, etc. color space  all of which are synthetically derived from mathematical models of human perception. Colors can exist in a scene that exist outside of the limitations but also it is possible for a camera to capture and a color space to contain  colors and tone differences that even  the healthiest of unaided human visual systems (Camera eye and visual cortex - AKA "naked eye") cannot see. A simple example of this is the rendering of the blend of evanescent colors recorded during a very long exposure.

"...but the colors will be represented within the color space. "

Not if you choose a color space, like sRGB that is smaller than the gamut of colors present in the scene. The colors in the scene that are outside of the color spaces gamut are automatically clamped (Joseph Holmes' term) to the limits of the color space. Equally important their relationship to colors that do fit into your chosen color space's gamut are distorted. (the architecture of the color space may also distort these naked eye perceived color relationships.

"If you want to maintain the closest relationship between the scene and the image using absolute colorimetric, then you must convert the camera data from RAW into a color space big enough to hold all the colors. For many people this is Adobe 1998. But that is only done for product shots (absolute colorimetric, I mean.)"

I do not know enough about using the Absolute Colorimetric rendering intent to comment. I do a fair amount of actual product and people photography , as well as warchitectural and lanatural world work. The vast majority of man made products fit nicely into Adobe RGB(1998) if not sRGB. The main problems I have with small (sRGB) and relatively large (Adobe RGB(1998)) color spaces clamping colors are outdoor scenes.

"The histogram only shows colors that are available to it - which is the monitor color space usually close to sRGB - so no matter how much you stare at it you aren't going to see any colors beyond the small sRGB color space of your monitor, which hopefully is calibrated."
Others have answered you on this but only partially. Which histogram are you referring to: the one you see on a camera or the one you are using in a raw processing program (Adobe Camera Raw, Aperture, Bibble, CaptureOne, Lightroom, etc.) or post-processing program (Photoshop, etc.)? If the camera, the histogram is determined by how you have set the camera up to do the raw to JPEG processing: primarily the color space setting, but also white balance, contrast, and the various "camera styles" settings.
If you are referring to the histogram in Adobe Camera Raw or in Photoshop  your statement is only true if you have chosen your display's profile to work within. While Lightroom primarily uses Adobe Camera Raw as the processing engine in the develop module ,  the histogram is based on what has been nicknamed "Melissa RGB" a variant of ProPhoto RGB with a TRC (Tone Response Curve) AKA "Gamma" of 2.2 as opposed to 1.8 as used in ProPhoto RGB. My understanding from both verbal and email interviews and communications with various Lightroom team members is that this was done to make it a better fit to the naked eye.

"If you start with Adobe 1998 color space, the CMS converts it into monitor color space by the conversion method you have chosen (perceptual, relative colorimetric or absolute colorimetric.) "

This is only true of the version of the document (photograph) that is sent to  the display.

"If you don't choose absolute colorimetric then no color data is clipped, only changed to fit what the monitor is able to show - that is good greens and blues and not so good reds and yellows."

Again I must pass on the discussion of the Absolute Colorimetric rendering intent.   As to what colorss the display is capable of displaying - that depends specifically of the characteristics of the  display being discussed and also  the settings you have used to calibrate it to and the device and software you are using to profile it with.

"Remember all you see is a representation of the colors that the monitor can produce, not what the printer can print."


That is absolutely true. The display and printer (really the specific combination of printer, ink (if ink is used) and paper) are two different devices. If you are printing the real proof is the print.


"Digital cameras don't have to deal with such a wide range of color - except in product shots in the studio, so sRGB is usually okay as a capture color space."

In my experience and also according to Joe Holmes (http://www.josephholmes.com/propages/AboutRGBSpaces.html) that is absolutely untrue.

I do not suffer from the illusion that my knowledge is infinite, especially regarding the science of color and how our photographic and reprographic devices render it , and am always looking to expand what I know. If you can help me in that process I am grateful.



« Last Edit: September 12, 2011, 09:25:41 am by Ellis Vener »
Logged
Pages: 1 [2]   Go Up