Pages: [1]   Go Down

Author Topic: Color Spaces and 'numbers'  (Read 8160 times)

Jeremy Payne

  • Guest
Color Spaces and 'numbers'
« on: April 19, 2009, 09:07:59 pm »

It took me a while, but I've finally pretty much gotten my head around color spaces and device profiles, etc.

Here's what I still don't understand.  Let's just take PhotoPro vs. sRGB for now ... I understand that PhotoPro is "bigger"/"wider" ...

... but when you represent them digitally, you still use the same 8-bit (or 16-bit) topography.

8-Bit PhotoPro and 8-Bit sRGB have the same number of distinct colors, right?  

Does thus just mean the "distance" between distinct 1 bit changes is "bigger" in PhotoPro?

But in a "continuous"/"analog" counting of colors, PhotoPro would have more colors than sRGB, right?

Help!
Logged

Damo77

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 69
    • http://www.bellephotography.net.au/
Color Spaces and 'numbers'
« Reply #1 on: April 19, 2009, 09:50:54 pm »

Yep, I'd say your understanding is pretty sound.
Logged
Damien

howardm

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1984
Color Spaces and 'numbers'
« Reply #2 on: April 20, 2009, 08:13:49 am »

here's an interesting chart (and website)

http://www.brucelindbloom.com/WorkingSpaceInfo.html

I think your understanding is correct but I'm sure more experienced people will chime in.

that's the reason 8 bit ProPhoto isn't desirable; the change in color for a 1 bit change can be
too large and can lead to posterization and quantization issues.

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2005
    • http://www.guillermoluijk.com
Color Spaces and 'numbers'
« Reply #3 on: April 20, 2009, 09:14:38 am »

That is correct, and the reason why not always the widest profile is better. The wider colour gammut a profile has, the poorer the gradations between different colours it will have, and this will make it less robust against strong postprocessing.


digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Color Spaces and 'numbers'
« Reply #4 on: April 20, 2009, 09:19:45 am »

The most saturated green we can define using 24 bit RGB is zero red, 255 green, zero blue. That color is different in sRGB versus Adobe RGB (1998). If you look at the CIE chromaticity diagram, it shows a 2D plot of colors we humans can see (based on experiments to build this theoretical color space). You would see that this green falls in a different location yet they have the same numbers. The difference is the scale. The big mongo scale or color space is that of human vision, that horseshoe shaped plot. G255 in sRGB is a different color than G255 in Adobe RGB (1998) as plotted on this graph and as defined as we humans would see that color. So the differences are the scale of the color space.

Or think of it another way. Numbers without a scale are useless. If you ask me the mileage I get on my new car and I simply say "32", you don't know if that's 32 miles per gallon or 32 miles or kilo. The gas is the same, the scale is different. In the example above, the colors are not the same yet the numbers are. A color space gives these numbers a defined color appearance based on how we see those colors and defines their scale within human vision.

ProPhoto doesn't have more colors than sRGB if we are using 24 bits to define the color space. The scale however is quite different. And since ProPhoto RGB has a wider gamut, the spaces (distance) between the 16.7 million colors in a 24 bit definition are wider. Think of sRGB's 16.7 million colors to be 16.7 million dots painted on a half inflated balloon. Think of ProPhoto as having the same number of dots, but this balloon is now twice as large. The distance between each dot is farther apart.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Jeremy Payne

  • Guest
Color Spaces and 'numbers'
« Reply #5 on: April 20, 2009, 09:34:08 am »

I love this stuff ... Ok .. so here's a question:

Say I take a RAW file and Import into Lighroom, process and then export to a 16-bit TIFF in Adobe RGB ... and then ...

Take the same RAW file, Open with ACR, process with the same settings as above and save as a 16-bit TIFF in Adobe RGB.

Does PhotoPro come into play in the former and not the latter because it is the native working space of Lightroom or are the two TIFFs absolutely identical ... or both ... or something else?
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Color Spaces and 'numbers'
« Reply #6 on: April 20, 2009, 09:39:01 am »

Quote from: Jeremy Payne
Say I take a RAW file and Import into Lighroom, process and then export to a 16-bit TIFF in Adobe RGB ... and then ...

Take the same RAW file, Open with ACR, process with the same settings as above and save as a 16-bit TIFF in Adobe RGB.

Does PhotoPro come into play in the former and not the latter because it is the native working space of Lightroom or are the two TIFFs absolutely identical ... or both ... or something else?

If the two app's are at parity in terms of version, the results are identical. They share the same processing engine.
« Last Edit: April 20, 2009, 09:39:19 am by digitaldog »
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Jeremy Payne

  • Guest
Color Spaces and 'numbers'
« Reply #7 on: April 20, 2009, 09:47:48 am »

Quote from: digitaldog
If the two app's are at parity in terms of version, the results are identical. They share the same processing engine.
Gotcha.  Makes sense - then that means the native working space of ACR prior to save or opening in Photoshop is also PhotoPro like Lightroom.

I would imagine that by staying in 16-bit as long as possible and being mindful of pushing into "alien" territory would more than compensate for the "bigger jumps" between bits in PhotoPro and why now many people recommend it/use it ... would that be fair to say?

Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Color Spaces and 'numbers'
« Reply #8 on: April 20, 2009, 09:55:30 am »

Quote from: Jeremy Payne
I would imagine that by staying in 16-bit as long as possible and being mindful of pushing into "alien" territory would more than compensate for the "bigger jumps" between bits in PhotoPro and why now many people recommend it/use it ... would that be fair to say?

Yes although when originally designed by Kodak, many years ago, not all functionality for doing a lot of high bit processing was available in Photoshop and Kodak did demonstrate the space was effective and useable in 8-bit. The comment about banding I suppose is absolutely possible depending on the document and edits. And I see little reason to toss away the extra data so I always archive the 16-bit TIFFs in ProPhoto. That said, I don't know that an 8-bit iteration would affect climate change or hurt the final output if push comes to shove.
« Last Edit: April 20, 2009, 09:55:56 am by digitaldog »
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

madmanchan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2115
    • Web
Color Spaces and 'numbers'
« Reply #9 on: April 20, 2009, 01:05:30 pm »

Jeremy, like Andrew said, LR and CR both use the ProPhoto RGB primaries internally. You'd get the same results both ways. To minimize quantization effects, LR and CR use at least 16 bits internally for color processing. I have not seen a case when using the ProPhoto RGB primaries with >= 16-bit math was the bottleneck in image quality; I always get quantization effects on the display first, rather than in the actual image.
Logged
Eric Chan

teddillard

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 717
    • http://www.teddillard.com
Color Spaces and 'numbers'
« Reply #10 on: April 23, 2009, 06:49:21 am »

If I may...  none of this stuff really made any sense to me until I started tinkering around with ColorThink.  www.chromix.com/colorthink/

Colorthink lets you look at gamut maps, but, more importantly, lets you look at gamut maps of image files as well as profiles and spaces.  It answers a whole crap-ton of questions, in a very visual way.  

One of the things I did at the outset was to make an image with just three colors- three pixels, and load it into ColorThink.  From there I could actually see the colors moving around, and their relationship changing, as they get converted, re-mapped, etc.  For instance.  Want to know what the difference is between perceptual intent and colormetric?  You can see very clearly by doing some conversions, and looking at your colors in ColorThink.  Want to see the difference between a file mapped into ProPhoto and the same file in sRGB?  You get the idea...  

Here's a little example...  I always heard that a JPG is, by nature, limited to sRGB, and that it doesn't make sense to work with JPG files in AdobeRGB.  One quick little test proved that "busted".  I made a file with a color out of gamut in sRGB, saved it as a JPG, and saw the color, still there, in the JPG, still outside of sRGB.  

Logged
Ted Dillard

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Color Spaces and 'numbers'
« Reply #11 on: April 23, 2009, 09:03:57 am »

Quote from: digitaldog
ProPhoto doesn't have more colors than sRGB if we are using 24 bits to define the color space. The scale however is quite different. And since ProPhoto RGB has a wider gamut, the spaces (distance) between the 16.7 million colors in a 24 bit definition are wider. Think of sRGB's 16.7 million colors to be 16.7 million dots painted on a half inflated balloon. Think of ProPhoto as having the same number of dots, but this balloon is now twice as large. The distance between each dot is farther apart.

This discussion raises some interesting points. With a bit depth of 8, both sRGB and ProPhotoRGB have 16.7 million encodable "colors". With sRGB all of these colors lie within the gamut of human vision, whereas some of the ProPhotoRGB colors lie beyond the gamut of human vision. Therefore, sRGB has more encodable values within the range of human vision. Since color is a perception, differences in encoded color values that can not be distinguished by humans are not really discrete colors. The standard for expressing the difference in color that can be perceived is Delta E (ΔE) and the volume of a color space can be expressed in terms of ΔE^3. Bruce Lindbloom lists the volume of various color spaces on his web site under the tag of information about RGB working spaces.

sRGB has a volume of 832,870 ΔE^3 whereas ProPhoto has a volume of 2,879,568 ΔE^3, and the number of perceivable colors in both spaces is far less than 16.7 million. ProPhotoRGB has more colors than sRGB, but whether or not one should use ProPhotoRGB depends on what colors are actually recordable in the scene. For most natural scenes, the gamut of ProPhotoRGB is unnecessarily large. Lindbloom has devised a BetaRGB that covers real world scenes and Gernot Hoffmann (warning: large PDF) has a plot of real world surface colors that are important to reproduce in photographic images. The area enclosed by this plot is larger than aRGB and smaller then ProPhotoRGB.

Choosing a "just right" space for each image is impractical, and IMHO 16 bit ProPhotoRGB is the best choice for general use, at least for images rendered with ACR. If ACR could render into an arbitrary color space, BetaRGB would be a reasonable choice.

Bill


Logged

Jeremy Payne

  • Guest
Color Spaces and 'numbers'
« Reply #12 on: April 23, 2009, 11:55:35 am »

Well ... this just crossed from theory into practice.

I got a "new"/refurbished Samsung XL20 last night for the very nice price of $675 ... had been planning on buying an HP 2475w, but this came along and I bit ...

Previously, I'd been editing on a Thinkpad (~40% Adobe gamut) and proofing on a wide-gamut (WG-CCFL) LCD HDTV (~90+% Adobe98).  I calibrate with a SpyderPro3.  I'll continue to use the Thinkpad - running 64-Bit Vista, LR 2.3 and Photoshop - and hook up the XL20 as a second monitor when needed.

After calibrating the XL20 and the laptop (again), I was amazed at some of the colors that popped out.  Some of the "wrong" colors on the laptop look "better" than some of the "right" colors on the XL20 ... but that's probably because I've been staring at the "wrong" colors for so long.

"sRGB Mode" is easy to use and does make non-color managed application more pleasant to use.

One question ... is the Spyder3Pro and utility a good hardware/software combo to calibrate a wide-gamut LED monitor?
Logged

Wayne Fox

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4237
    • waynefox.com
Color Spaces and 'numbers'
« Reply #13 on: April 23, 2009, 03:49:49 pm »

Quote from: bjanes
This discussion raises some interesting points. With a bit depth of 8, both sRGB and ProPhotoRGB have 16.7 million encodable "colors". With sRGB all of these colors lie within the gamut of human vision, whereas some of the ProPhotoRGB colors lie beyond the gamut of human vision. Therefore, sRGB has more encodable values within the range of human vision. Since color is a perception, differences in encoded color values that can not be distinguished by humans are not really discrete colors. The standard for expressing the difference in color that can be perceived is Delta E (ΔE) and the volume of a color space can be expressed in terms of ΔE^3. Bruce Lindbloom lists the volume of various color spaces on his web site under the tag of information about RGB working spaces.

sRGB has a volume of 832,870 ΔE^3 whereas ProPhoto has a volume of 2,879,568 ΔE^3, and the number of perceivable colors in both spaces is far less than 16.7 million. ProPhotoRGB has more colors than sRGB, but whether or not one should use ProPhotoRGB depends on what colors are actually recordable in the scene. For most natural scenes, the gamut of ProPhotoRGB is unnecessarily large. Lindbloom has devised a BetaRGB that covers real world scenes and Gernot Hoffmann (warning: large PDF) has a plot of real world surface colors that are important to reproduce in photographic images. The area enclosed by this plot is larger than aRGB and smaller then ProPhotoRGB.

Choosing a "just right" space for each image is impractical, and IMHO 16 bit ProPhotoRGB is the best choice for general use, at least for images rendered with ACR. If ACR could render into an arbitrary color space, BetaRGB would be a reasonable choice.

Bill

Is there any real perceptible advantage to having a just right space?  Is there any real perceptible advantage to using a smaller space than ppRGB that can still contain the image data .. especially when working with 16bit files?  would using a "betaRGB" space have any real advantage?  

I haven't seen any real issues using ppRGB/16bit.  I see others complain of issues that they claim are solved by moving to a different working space, such as sRGB, but I'd be willing to bet ppRGB wasn't the actual cause of the issue, moving to sRGB just masked it so the assumption is made the working space was the cause.

I guess I can see the logic in theory, but I'm not sure there is a real practical need.  I also wonder if sometimes we exceed "real world" colors in our conversion that still produce a look we like and is still realistic  - colors that would  be clipped in this "betaRGB" space concept.

Logged

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Color Spaces and 'numbers'
« Reply #14 on: April 23, 2009, 06:04:26 pm »

Quote from: Wayne Fox
Is there any real perceptible advantage to having a just right space?  Is there any real perceptible advantage to using a smaller space than ppRGB that can still contain the image data .. especially when working with 16bit files?  would using a "betaRGB" space have any real advantage?  

I haven't seen any real issues using ppRGB/16bit.  I see others complain of issues that they claim are solved by moving to a different working space, such as sRGB, but I'd be willing to bet ppRGB wasn't the actual cause of the issue, moving to sRGB just masked it so the assumption is made the working space was the cause.

I guess I can see the logic in theory, but I'm not sure there is a real practical need.  I also wonder if sometimes we exceed "real world" colors in our conversion that still produce a look we like and is still realistic  - colors that would  be clipped in this "betaRGB" space concept.

If you are working with 16 bit files, I doubt that there would be any advantage to using betaRGB. Like you, I haven't had any problems with ProPhotoRGB. If you could get by using 8 bits, betaRGB or a just the right size space might help prevent posterization.

Bill
Logged

Wayne Fox

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4237
    • waynefox.com
Color Spaces and 'numbers'
« Reply #15 on: April 23, 2009, 09:39:16 pm »

Quote from: bjanes
If you are working with 16 bit files, I doubt that there would be any advantage to using betaRGB. Like you, I haven't had any problems with ProPhotoRGB. If you could get by using 8 bits, betaRGB or a just the right size space might help prevent posterization.

Bill

from that perspective I can see where it may make a difference.
Logged
Pages: [1]   Go Up