Ad
Ad
Ad
Pages: [1] 2 »   Bottom of Page
Print
Author Topic: RGB Working space  (Read 9042 times)
allan67
Jr. Member
**
Offline Offline

Posts: 69


WWW
« on: September 01, 2006, 04:40:52 PM »
ReplyReply

Hello,

If I understand the process correctly, when an image in PS gets printed, the WHOLE working space gets compressed (or clipped) to the printer gamut. That means that when using Perceptual intent even colours that are within the printer gamut also get moved (to a different extent), depending on the size of the working space - the larger it is, the more colours are moved to accomodate the far-away colours at the border of working space. With Relative intent the out-of-gamut colours are clipped to the border of printer gamut and colours within it don't get moved at all.

So, it appears that to minimise the shift of colours within the gamut one needs to use as small a working space as possible that can still contain all the colour of the image. (J. Holmes proposes such colour spaces on his website.)

The question is, how do I know if the given working space is large enough to contain all colours of the RAW image? Is there a way to get the gamut of colours in RAW file and compare it to the gamut of working space?

Allan
Logged
Stephen Best
Guest
« Reply #1 on: September 01, 2006, 08:28:54 PM »
ReplyReply

Quote
If I understand the process correctly, when an image in PS gets printed, the WHOLE working space gets compressed (or clipped) to the printer gamut.
[a href=\"index.php?act=findpost&pid=75263\"][{POST_SNAPBACK}][/a]

No. The working space doesn't get compressed. The image colours are instead mapped directly to their absolute coordinates in the reference space, and these in turn are mapped/compressed to fit the output space. The mapping/compression from the reference space (PCS) is fixed and is defined when the output profile is built. So it makes no difference to output what working space is used. The output tables have to handle all colours in the PCS, though the profiling package may make assumptions about the typical input space when building its perceptual tables.

However .... you should always use the smallest working space that retains the image colours you want. To determine what space you typically need, bring your raw image in in a large space (say ProPhoto RGB) and set the soft-proofing space in turn to sRGB, Adobe RGB etc. Turn on the gamut warning to see which image colours (if any) would have got clipped had you brought the image in in that smaller space. Some colours you may not care whether they're clipped or not. Map the space to the image, don't just blindly use ProPhoto RGB as some recommend.
Logged
allan67
Jr. Member
**
Offline Offline

Posts: 69


WWW
« Reply #2 on: September 02, 2006, 11:48:17 AM »
ReplyReply

Quote
However .... you should always use the smallest working space that retains the image colours you want.
[a href=\"index.php?act=findpost&pid=75277\"][{POST_SNAPBACK}][/a]

Stephen, thank you for the answer and tip on checking the real gamut of the image.

If only the colours that are really in the image get shifted into the output space, why is it necessary to use the smallest possible working space?
I thought that was exactly because the whole working space was compressed...  

Allan
Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2822



« Reply #3 on: September 02, 2006, 12:33:29 PM »
ReplyReply

Quote
If only the colours that are really in the image get shifted into the output space, why is it necessary to use the smallest possible working space?
I thought that was exactly because the whole working space was compressed...  
[a href=\"index.php?act=findpost&pid=75313\"][{POST_SNAPBACK}][/a]

Allan's misconception is very common and I was under that impression too until I recently participated in a thread with Stephen. The smaller space has better coding efficiency, but if you use a bit depth of 16 with ProphotoRGB, this should not be an issue.

If you use Adobe Camera Raw, you can see if the color space is large enough by looking for clipping in the histograms. If there is none, your space is large enough. If clipping is evident, use a larger space.

Bill
Logged
opgr
Sr. Member
****
Offline Offline

Posts: 1125


WWW
« Reply #4 on: September 02, 2006, 01:11:26 PM »
ReplyReply

Quote
If only the colours that are really in the image get shifted into the output space, why is it necessary to use the smallest possible working space?

Advantages of a larger color space:
- It may be able to represent all the colors that are captured where a smaller space may not,
- in case of ACR + ProPhoto RGB, there is a more direct connection between ACR internals and the colorspace which translates to theoretically less loss in conversions.

Disadvantages of a larger color space:
- less coding efficiency for pastel colors,
- coarser control over color corrections.
(Skintones would be the obvious example.)

It's primarily the color control issue that would suggest the use of the best fitting colorspace.
Logged

Regards,
Oscar Rysdyk
theimagingfactory
Stephen Best
Guest
« Reply #5 on: September 02, 2006, 05:03:44 PM »
ReplyReply

Quote
If only the colours that are really in the image get shifted into the output space, why is it necessary to use the smallest possible working space?
[{POST_SNAPBACK}][/a]

As others have pointed out, the main problem with ProPhoto RGB is inefficient use of the volume and the concomitant difficulty of doing fine adjustments to image colours etc. Photoshop's editing controls weren't really designed to do the adjustments in a space as large as ProPhoto RGB with the granularity one requires. But try it for yourself. If, like me, you find doing final tweaks to an image in ProPhoto RGB frustrating you can always downsize the space to something more reasonable. It's not so much about retaining theoretical colours you may or may not see in the print, but more about how practical it is to actually get the final print you want. You may want to read this pointed article:

[a href=\"http://www.jeremydaalder.com/singleArticle.php?articleID=6]http://www.jeremydaalder.com/singleArticle.php?articleID=6[/url]

I use Beta RGB for most of my own work (from scanned trannies) but this is still fairly large. It's a better match for Fujichrome than Joseph Holmes' spaces. More info on Beta RGB here:

http://www.brucelindbloom.com/BetaRGB.html
Logged
Hendrik
Jr. Member
**
Offline Offline

Posts: 93


WWW
« Reply #6 on: September 03, 2006, 02:54:07 AM »
ReplyReply

One thing you should keep in mind when you choose the smallest color space (as I also do), is when you going to use filters or edits that boost the saturation. You can get clipping when you choose a tight fitting color space.

It’s wise to have an idea how you want to edit your picture before you start. It’s a matter of making wise decisions and that means sometimes you select sRGB, sometimes Adobe RGB and sometimes ProPhoto RGB.

You can always convert to a larger color space when you need it, but before you get clipping. Clipping means, …data lost.
Logged
standard_observer
Newbie
*
Offline Offline

Posts: 11


« Reply #7 on: September 03, 2006, 04:07:00 AM »
ReplyReply

Quote
.... Photoshop's editing controls weren't really designed to do the adjustments in a space as large as ProPhoto RGB with the granularity one requires. But try it for yourself. If, like me, you find doing final tweaks to an image in ProPhoto RGB frustrating you can always downsize the space to something more reasonable.
But do you really see this effect in practice
(with ProPhoto RGB at 16 bit/ch)?

The editable tonal scale from RGB 0 to 255 is the same in all working spaces; ranging from L* 0 to 100 in absolute terms. Just the in-between values differ depending on working space gamma.

Also the 360 shades of accessible hues are roughly the same with all common working spaces. In detail, the absolute meaning of course depends on the CIE xy coordinates of the gamut primaries (their angle around the white point).  But it is not so that smaller spaces would offer a finer control. Not at all.

The only main difference is along the saturation axis from 0 to 100 according HSB readings or the Hue&Sat.-tool.  As HSB is derived from RGB, one step (%) is more ‘Lab’-effective in a large working space compared to a small one.

However, there’s always the option to reduce the Opacity of the adjust layer, in order to make finer saturation changes - if needed.

--
Logged
Stephen Best
Guest
« Reply #8 on: September 03, 2006, 04:44:47 AM »
ReplyReply

Quote
But do you really see this effect in practice
(with ProPhoto RGB at 16 bit/ch)?
[a href=\"index.php?act=findpost&pid=75351\"][{POST_SNAPBACK}][/a]

The bit depth isn't so much the issue as the granularity of change you can effect with current Photoshop controls. Compare fine tuning skin tones in sRGB and ProPhoto RGB. Reducing opacity of your Curves etc. adjustment layer to something like 50% is an effective measure ... but if you look at the differences you actually manage to realize in the print you may be asking why you bothered with ProPhoto RGB in the first place. I'm not saying ProPhoto RGB isn't required for some subjects, but you should be aware of why you're using it and what the downsides are.
« Last Edit: September 03, 2006, 04:46:12 AM by Stephen Best » Logged
Stephen Best
Guest
« Reply #9 on: September 03, 2006, 08:26:22 AM »
ReplyReply

Quote
The only main difference is along the saturation axis from 0 to 100 according HSB readings or the Hue&Sat.-tool.  As HSB is derived from RGB, one step (%) is more ‘Lab’-effective in a large working space compared to a small one.
[a href=\"index.php?act=findpost&pid=75351\"][{POST_SNAPBACK}][/a]

You're right in that the saturation axis is the main issue, but you forget that Curves adjustments also affect saturation. In fact, the better I get at Curves the less need I have for the Hue/Saturation tool.

Here's some numbers to highlight the problem:

I just created a new image in sRGB (16-bit to minimize rounding) and filled it with an arbitrary colour ... in this instance 225,163,231. The Lab coordinates of this are 75,32,-25. Then I did a curve adjustment by placing a single random point (input 230, output 235) on just the red channel. This changed the colour to 75,34,-24. I then reverted back to the original colour, converted the image to ProPhoto RGB and applied exactly the same curve adjustment. The resultant colour is now 76,38,-23 ... as you'd agree (and expect) a much bigger change. In dE2000 terms, the difference has increased from 1.23 to 3.25. There's bound to be some rounding error in this but the fact is that adjustments in larger spaces just result in larger adjustments. Now you can just make smaller adjustments in larger spaces to compensate but if the adjustment you're trying to achieve is subtle, you may find this difficult. I wouldn't be surprised if a future version of Photoshop addressed exactly this issue.
« Last Edit: September 03, 2006, 08:32:55 AM by Stephen Best » Logged
opgr
Sr. Member
****
Offline Offline

Posts: 1125


WWW
« Reply #10 on: September 03, 2006, 08:50:08 AM »
ReplyReply

Quote
But do you really see this effect in practice
(with ProPhoto RGB at 16 bit/ch)?


While I agree with you that this is largely an academic issue, you might also want to consider near-neutral gradations. For example, the subtly toned metallic silver grays of cars that have been popular the last couple of years. A car advertisement would be filled with such gradations and the more tones are available to build that gradation, the better. And that's BEFORE you even do any corrections...
Logged

Regards,
Oscar Rysdyk
theimagingfactory
allan67
Jr. Member
**
Offline Offline

Posts: 69


WWW
« Reply #11 on: September 03, 2006, 11:21:56 AM »
ReplyReply

Thanks a lot to everybody paticipating here.
Now I'm starting to understand the WHY's and not only HOW's of colour correction in different work spaces.
Most informative (for me, at least)

Allan
Logged
OutsideShooter
Newbie
*
Offline Offline

Posts: 39


« Reply #12 on: September 03, 2006, 06:22:35 PM »
ReplyReply

Quote
However .... you should always use the smallest working space that retains the image colours you want. To determine what space you typically need, bring your raw image in in a large space (say ProPhoto RGB) and set the soft-proofing space in turn to sRGB, Adobe RGB etc. Turn on the gamut warning to see which image colours (if any) would have got clipped had you brought the image in in that smaller space. Some colours you may not care whether they're clipped or not. Map the space to the image, don't just blindly use ProPhoto RGB as some recommend.
[a href=\"index.php?act=findpost&pid=75277\"][{POST_SNAPBACK}][/a]

Pardon me for getting lost in the peripheral discussion, but Alan's inquiry was: The question is, how do I know if the given working space is large enough to contain all colours of the RAW image? Is there a way to get the gamut of colours in RAW file and compare it to the gamut of working space?

Allan

Then basically your answer, Stephen,  was: To determine what space you typically need, bring your raw image in in a large space (say ProPhoto RGB) and set the soft-proofing space in turn to sRGB, Adobe RGB etc. Turn on the gamut warning to see which image colours (if any) would have got clipped had you brought the image in in that smaller space.

Now what I'm asking is, how does one find the steps you are assigning? Can you give us a step by step? I, for one, have no idea how to find the soft-proofing space not the gamut warning.

Thanks

Rich
« Last Edit: September 03, 2006, 06:23:26 PM by OutsideShooter » Logged

Stephen Best
Guest
« Reply #13 on: September 03, 2006, 07:06:05 PM »
ReplyReply

Quote
Now what I'm asking is, how does one find the steps you are assigning? Can you give us a step by step? I, for one, have no idea how to find the soft-proofing space not the gamut warning.
[a href=\"index.php?act=findpost&pid=75417\"][{POST_SNAPBACK}][/a]

Obviously you're not going to do this for every image. The intention is to get a feel for working spaces and how they might match up with your typical images. The graph in ACR will just show you that some colours may or may not be clipped, not whether they're actually worthwhile retaining.

Here's the detailed procedure. After you've converted the RAW image to ProPhoto RGB and have it open in Photoshop, go to the View menu and select Proof Setup then Custom. Select sRGB (or sRGB IEC61966-2.1) from the Device to Simulate dropdown. The other settings don't matter. Your display will then show roughly what the image would look like had it instead been converted to sRGB. Probably not much changed as your display is likely limited to something like sRGB anyway. You can then turn on the Gamut Warning from the View menu to show the clipped colours, namely those outside of the smaller space. It helps to set the Gamut Warning colour to something like a bright green (0,255,0) in Photoshop preferences. Repeat the above for other spaces (Adobe RGB 1998 etc.). Basically, all you're doing here is soft-proofing the image, but you can use working or display spaces just the same as printer spaces. Whether the colours you see (or rather don't see) will be retained in the print is another matter.
Logged
David White
Sr. Member
****
Offline Offline

Posts: 272



WWW
« Reply #14 on: September 03, 2006, 10:10:19 PM »
ReplyReply

Quote
I just created a new image in sRGB (16-bit to minimize rounding) and filled it with an arbitrary colour ... in this instance 225,163,231. The Lab coordinates of this are 75,32,-25. Then I did a curve adjustment by placing a single random point (input 230, output 235) on just the red channel. This changed the colour to 75,34,-24. I then reverted back to the original colour, converted the image to ProPhoto RGB and applied exactly the same curve adjustment. The resultant colour is now 76,38,-23 ... as you'd agree (and expect) a much bigger change. In dE2000 terms, the difference has increased from 1.23 to 3.25. There's bound to be some rounding error in this but the fact is that adjustments in larger spaces just result in larger adjustments. Now you can just make smaller adjustments in larger spaces to compensate but if the adjustment you're trying to achieve is subtle, you may find this difficult. I wouldn't be surprised if a future version of Photoshop addressed exactly this issue.
[{POST_SNAPBACK}][/a]
All this test does is demonstrate that RGB values map differently in different color spaces.  Did you notice that when you reverted back to the original color and then converted the image to ProPhoto that the RGB values changed to 197,157,217?  To compound the problem you then made RGB value adjustments which do not map the same in a different color space.

Take a look at [a href=\"http://brucelindbloom.com/]Bruce Lindholm's CIE Color Calculator[/url] under the CALC section and try the numbers for yourself.
Logged

David White
Stephen Best
Guest
« Reply #15 on: September 03, 2006, 10:55:23 PM »
ReplyReply

Quote
Did you notice that when you reverted back to the original color and then converted the image to ProPhoto that the RGB values changed to 197,157,217?
[a href=\"index.php?act=findpost&pid=75434\"][{POST_SNAPBACK}][/a]

Yes, as you'd expect. The intention of the preceding was to show that a set curve adjustment will have a larger effect in the larger space ... that was all.
Logged
standard_observer
Newbie
*
Offline Offline

Posts: 11


« Reply #16 on: September 04, 2006, 04:47:51 PM »
ReplyReply

Quote
I just created a new image in sRGB (16-bit to minimize rounding) and filled it with an arbitrary colour ... in this instance 225,163,231. The Lab coordinates of this are 75,32,-25. Then I did a curve adjustment by placing a single random point (input 230, output 235) on just the red channel. This changed the colour to 75,34,-24. I then reverted back to the original colour, converted the image to ProPhoto RGB and applied exactly the same curve adjustment. The resultant colour is now 76,38,-23 ... as you'd agree (and expect) a much bigger change. In dE2000 terms, the difference has increased from 1.23 to 3.25. There's bound to be some rounding error in this but the fact is that adjustments in larger spaces just result in larger adjustments. Now you can just make smaller adjustments in larger spaces to compensate but if the adjustment you're trying to achieve is subtle, you may find this difficult. I wouldn't be surprised if a future version of Photoshop addressed exactly this issue.
[a href=\"index.php?act=findpost&pid=75360\"][{POST_SNAPBACK}][/a]
Actually I think that your well-elaborated example leaves the door open for individually different conclusions:

With this curve on the R channel, the average deltaE per increment is 3.25/(325-320)= 0.65.  Accordingly, with a minimal curve of input/output= 230/231 (in ProPhoto RGB) the resulting change is just at the border of (my) perception.

So I’m afraid that for my purposes this seems to be enough precision.

Alternatively, if you repeat the test but now applying the curve (230/235) on the RGB composite channel, the resulting deltaE2000 is almost the same: 2.68 in sRGB vs 2.53 in pRGB – with ProPhoto RGB having an irrelevant competitive edge with regard to precision (due to its lower local gamma).


Quote
While I agree with you that this is largely an academic issue, you might also want to consider near-neutral gradations. For example, the subtly toned metallic silver grays of cars that have been popular the last couple of years. A car advertisement would be filled with such gradations and the more tones are available to build that gradation, the better. And that's BEFORE you even do any corrections...
[a href=\"index.php?act=findpost&pid=75363\"][{POST_SNAPBACK}][/a]

Please kindly explain.

--
Logged
Stephen Best
Guest
« Reply #17 on: September 04, 2006, 05:05:01 PM »
ReplyReply

Quote
Alternatively, if you repeat the test but now applying the curve (230/235) on the RGB composite channel, the resulting deltaE2000 is almost the same: 2.68 in sRGB vs 2.53 in pRGB – with ProPhoto RGB having an irrelevant competitive edge with regard to precision (due to its lower local gamma).
[a href=\"index.php?act=findpost&pid=75489\"][{POST_SNAPBACK}][/a]

Tonality isn't the issue, saturation is. Maybe it was a flawed example, but nobody who has edited images in both sRGB and ProPhoto RGB for colour would argue that the level of control is comparable. The reason I gave numbers is because I get the impression that most here would rather argue numbers than trust what their eyes tell them.
Logged
Jonathan Wienke
Sr. Member
****
Offline Offline

Posts: 5759



WWW
« Reply #18 on: September 05, 2006, 11:41:58 AM »
ReplyReply

IMO the disadvantages of large color spaces like ProPhoto are not generally significant in the real world, as long as one stays in 16-bit mode. Yes there are differences in how some of the settings work depending on the color space you're working with, but these differences do not ever significantly impede's one's ability to accomplish an intended task. I've done weddings, yearbook portraits, family portraits, as well as landscape/fine art work in ProPhoto, and have never encountered a situation where Photoshop's color controls were insufficiently precise for anything I wanted to do. I do most color correction in Camera RAW; I have it calibrated to my cameras, and in most cases all I have to do is set the white balance appropriately, and all the colors fall in to place without significantly tweaking in Photoshop itself. However, I do a lot of toned monochrome, and blended toned monochrome + color, and the Hue/Saturation, Levels, and Curve dialogs have never annoyed me with their imprecision while using ProPhoto. I have to use different Hue/Saturation dialog settings to achieve the same tone when working in ProPhoto instead of sRGB, but the choice of color space has never stopped me from achieving the effect I want. I edited this image in ProPhoto, and then downsampled to sRGB for web display after all other edits were complete.
Logged

bjanes
Sr. Member
****
Offline Offline

Posts: 2822



« Reply #19 on: September 05, 2006, 01:08:18 PM »
ReplyReply

Quote
IMO the disadvantages of large color spaces like ProPhoto are not generally significant in the real world, as long as one stays in 16-bit mode. Yes there are differences in how some of the settings work depending on the color space you're working with, but these differences do not ever significantly impede's one's ability to accomplish an intended task. I've done weddings, yearbook portraits, family portraits, as well as landscape/fine art work in ProPhoto, and have never encountered a situation where Photoshop's color controls were insufficiently precise for anything I wanted to do. I do most color correction in Camera RAW; I have it calibrated to my cameras, and in most cases all I have to do is set the white balance appropriately, and all the colors fall in to place without significantly tweaking in Photoshop itself. However, I do a lot of toned monochrome, and blended toned monochrome + color, and the Hue/Saturation, Levels, and Curve dialogs have never annoyed me with their imprecision while using ProPhoto. I have to use different Hue/Saturation dialog settings to achieve the same tone when working in ProPhoto instead of sRGB, but the choice of color space has never stopped me from achieving the effect I want. I edited this image in ProPhoto, and then downsampled to sRGB for web display after all other edits were complete.
[a href=\"index.php?act=findpost&pid=75582\"][{POST_SNAPBACK}][/a]

I agree fully with Jonathan here. With each version, Photoshop provides more comprensive support for 16 and higher bit depth. In 16 bit mode, one would expect that curves could be expressed in 16 bit terms, but apparently Adobe thinks that 8 bit precision is sufficient. I do note that many gurus such as Bruce Fraser do recommend ProPhotoRGB as a working space and have not mentioned the lack of precision as a significant disadvantage.

If you need more precision, you could use a curves layer and adjust the opacity, or you could even convert to sRGB and get precisely the same effect as if the file were orginally rendered into sRGB, at least with Adobe Camera Raw. However, the converse is not true--once out of gamut colors are clipped into a smaller space, they can not be recovered.
Logged
Pages: [1] 2 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad