Lee Varis' introduction to color management

[Doug Gray]

I agree with most of Andrew's points. In particular it's normal for RAW files to have large numbers of pixels that are outside of sRGB or even Adobe RGB. There are two main reasons. Reflective surfaces are only capable of having the most saturated colors when their overall reflected light is low. This can be counterintuitive but easily shown with the ColorChecker SG. Most of the many colors in it's patches that are outside of sRGB are quite dim. Also, shadows are often illuminated by light that has already been color cast in the same direction as the illuminated components in the shadows which increases the saturation of the reflected light. Also, there is more sensor noise in shadows. This increases that portion of colors that are inside sRGB or Adobe RGB since the noise induced pixels that are outside are clipped by RAW conversion This doesn't happen to any appreciable degree with ProPhoto.

That said, printer gamuts shrink a lot in shadows and at even moderately low L values their gamuts have a hard time even reaching that of sRGB.

Lee also discounts soft proofing. Now it is true that one needs to have a complete color management process to count on soft proofing. The only times I've had a soft proof fail is when something was broken in the color management cycle. That doesn't happen often but usually it's when I'm using a flawed canned profile. They have improved a lot over the years but some companies provide really awful profiles that have screwy RCs and worse, have A2B tables return Lab values that aren't even close to what the printer puts out. If one soft proofs with those profiles they will not get a good screen proof match to the print and it will sour them on the very valuable role soft proofing offers.

[Peter_DL]

http://www.adobe.com/digitalimag/pdfs/phscs2ip_colspace.pdf

…So the advantage of ProPhoto isn't only about retaining all those out-of-gamut colors it's also about maintaining the dissimilarities between them, so that you can map them into a printable color space as gradations rather than ending up as blobs.

The saturation clipping shown at page 11 of the pdf seems to support this point. The image has a gamut that gets clipped upon changing the output space in ACR from ProPhoto RGB to sRGB. I also often see this adverse effect when I want to output an image from ACR into sRGB to send it to a print service.

But where do these colors come from ?
Is it really from the scene ? No, more often:

Image editing on large ProPhoto-gamut basis in ACR/LR tends to create those saturated colors, colors which were not in scene, and which later on – upon changing the output space from ProPhoto RGB to sRGB - posterize at the border of sRGB.
And, to be clear, with image editing I do not mean to crank up the Saturation slider. It is simple tonal settings, even and in particular the default tone curve in ACR/LR, which can push colors out of sRGB.

I had provided an example earlier, here. The described effect is easy to reproduce. In ACR/LR, whenever the change of the output space from ProPhoto RGB to sRGB produces a bunch of saturation clipping, just eliminate the tone curve. This can be done by using a dng profile with a linear tone curve, or, by going to PV2010 to zero the tonal settings in the Basic tab, plus Point Curve linear. The saturation clipping will typically get greatly reduced.

ProPhoto RGB is at times like a self-fulfilling prophecy - its use creates its need.

Peter

[D Fosse]

Image editing on large ProPhoto-gamut basis in ACR/LR tends to create those saturated colors, colors which were not in scene

Yes, exactly the point I was trying to make. They're artifacts.

ProPhoto RGB is at times like a self-fulfilling prophecy - its use creates its need.

That's wonderfully put and crams the whole issue nicely into a nutshell. With a fitting analogy to go with it.

[Simon Garrett]

Image editing on large ProPhoto-gamut basis in ACR/LR tends to create those saturated colors, colors which were not in scene, and which later on – upon changing the output space from ProPhoto RGB to sRGB - posterize at the border of sRGB.
And, to be clear, with image editing I do not mean to crank up the Saturation slider. It is simple tonal settings, even and in particular the default tone curve in ACR/LR, which can push colors out of sRGB.

Yes, exactly the point I was trying to make. They're artifacts.

That's wonderfully put and crams the whole issue nicely into a nutshell. With a fitting analogy to go with it.

If you do the same editing Peter_DL suggests (applying a tone curve) but in sRGB instead of ProPhoto RGB then the same artificial colours are created, but you don't see them because they are clipped in sRGB, surely? 

I quite agree that tone changes (as well as saturation changes) can result in colour values that weren't in the original image.  But I don't think it's editing in ProPhoto RGB that creates the artifacts; it's the edit operation itself. 

However, by editing in ProPhoto RGB then the colours aren't clipped after every edit operation, but only once if/when the images is converted to sRGB.  I would have thought this would end up with fewer artifacts overall.

[D Fosse]

A curve adjustment will never hard clip. Levels will - you have to move the endpoint.

A "curves saturation" adjustment has been implemented in Photoshop, and it's called Vibrance. You get the same effect with an S-curve set to Color blend mode.

Edit: (of course, vibrance was in ACR/Lightroom before that.)

[digitaldog]

Is it really from the scene ?

I'm not sure, based on how capture devices record the scene if you will, we can answer that. We're dealing with output referred imagery in the cases under discussion here, certainly with Lee's video examples.
The proof is in the print too. The example's I've seen and have shown in my video are not attractive when using a smaller gamut working space upon the final print. For many, the final print is all that maters.

[BobDavid]

Slightly OT: Andrew Rodney, AKA Digital Dog, wrote something awhile back on this thread about the benefits of processing RAW files in Prophoto RGB.  I've generally processed RAWs in aRGB-16 bit. On occasion, if a native file showed a lot of saturated reds, I'd processed the RAW file in sRGB-16 bit. Until a few weeks ago, my principle monitor (an Eizo CG211) covered approximately 95% of the sRGB space.

I my MO involves processing RAWs in aRGB-16 bit for the master Tiff.  Using aRGB as a color matching starting point, seems to make sense based on my assumption that it was okay to bounce back and forth between aRGB and LAB without destroying data. After a couple of years, I got pretty good (not great) at interpreting numbers despite not being able to rely on a state-of-the-art Eizo sRGB monitor. The monitor represented luminance fine, and in soft proofing mode, the monitor represented print tonality reasonably well. Still, the monitor's inability to show anything beyond 95% sRGB was a very weak link in my color management chain.

Now that I've replaced the CG with a wide gamut aRGB monitor, soft proofing is way more accurate. Since Mr. Rodney recommends Prophoto as a default profile, I am still unable to see all of the data on the new Eizo aRGB monitor. The point he makes about the advantages of Prophoto is true, especially for fine art ink jet printing. I didn't realize how much subtle data got clipped when printing from an aRGB file. Blotchy and clipped shadow detail, particularly low key photos, is largely eliminated in Prophoto.

Until monitors are able to display Prophoto, I think color management is still an immature technology. I do however want to restate a question: Is bouncing back and forth from Prophoto to LAB a destructive process? I like LAB for many reasons and Prophoto for others.

[fdisilvestro]

If you believe that there are no colors outside of sRGB in nature and all you need for editing is sRGB, please do so. For me, it is just nonsense

[digitaldog]

Slightly OT: Andrew Rodney, AKA Digital Dog, wrote something awhile back on this thread about the benefits of processing RAW files in Prophoto RGB.

My main point was that IF you process raws in the ACR engine, you're using a ProPhoto RGB gamut. Other's can argue with Thomas Knoll about that option. But that's what's happening under the hood. That's true with rendered images too but if you're processing say an sRGB rendered image in ACR/LR, it's being processed in the ProPhoto RGB gamut.

I adapted to aRGB and Prophoto, by toggling on soft proofing and enabling the gamut preview.

The OOG overlay in both Photoshop and LR are buggy and inaccurate. Mostly totally unnecessary. The monitor gamut OOG in LR is useful!

The Out Of Gamut Overlay in Photoshop and Lightroom

In this 25 minute video, I'll cover everything you need to know about the Out Of Gamut (OOG) overlay in Photoshop and Lightroom. You'll see why, with a rare exception, you can ignore this very old feature and still deal with out of gamut colors using modern color management tools.

YouTube: https://www.youtube.com/watch?v=00O-GTDyL0w
High resolution: http://digitaldog.net/files/OOG_Video.mp4

Using aRGB as a color matching starting point, seemed to make sense based on my assumption that it was okay to bounce back and forth between aRGB and LAB without destroying data

IF you're converting, no, that's not at all the case! If you're using Lab readout's in Adobe RGB (1998), fine.

Needless the state-of-the-art Eizo sRGB monitor was inherently colorblind.

That will be the case for some time; data that's out of display gamut. So you have an option; clip data you can capture and output but can't see or keep that data and be very careful about how you edit very saturated colors. Tip: When working in such wide gamut spaces, if you're editing using say Vibrance or Saturation, at such a point you don't see the  preview update as you move the slider, BACK OFF! You're probably editing the data blindly.

Until monitors are able to display Prophoto, I assume color management is still an immature technology

That will never happen until we possibly evolve to the 'star child' status at the end of 2001 and our visual system can see outside the current limitations. IOW, ProPhoto RGB contains device values that are not colors as we can't see them.

Is bouncing back and forth from Prophoto to LAB a destructive process?

Oh yes! Every time a conversion to LAB is produced, the rounding errors and severe gamut mismatch between the two spaces can account for data loss, known as quantization errors. The amount of data loss depends on the original gamut size and gamma of the working space. For example, if the working space is Adobe RGB, which has 256 values available, converting to 8- bit LAB reduces the data down to 234 levels for neutrals. The net result is a loss of 22 levels. Doing the same conversions from ProPhoto RGB reduces the data to only 225 values, producing a loss of 31 levels. Bruce Lindbloom, a well-respected color geek and scientist, has a very useful Levels Calculator,which allows you to enter values to determine the actual number of levels lost to quantization (see the “Calc page”at http://www.brucelindbloom.com). But doing this on high-bit data, something Lee dismisses should reduce the data loss such, it's not an issue. But is it necessary?

[digitaldog]

Speaking of Lab readout's, it is somewhat interesting how Lightroom vs. Photoshop proper report the values when it finds an 'illegal' color, like G255 in ProPhoto RGB. It would be nice if the two worked the same. For that matter, be nice if Photoshop could toggle to the 0-100% scale just so people in LR could (within reason) use the same scale in PS if they wanted to. Be nice if the OOG overlay wasn't buggy and showed us a range of how far these colors are using maybe three 'strengths' of color. A tiny bit OOG, a moderate amount and (G255 in ProPhoto) not a color at all.

[BobDavid]

Speaking of Lab readout's, it is somewhat interesting how Lightroom vs. Photoshop proper report the values when it finds an 'illegal' color, like G255 in ProPhoto RGB. It would be nice if the two worked the same. For that matter, be nice if Photoshop could toggle to the 0-100% scale just so people in LR could (within reason) use the same scale in PS if they wanted to. Be nice if the OOG overlay wasn't buggy and showed us a range of how far these colors are using maybe three 'strengths' of color. A tiny bit OOG, a moderate amount and (G255 in ProPhoto) not a color at all.

I've never used LR. So LR does not use the same engine as ACR?

[digitaldog]

I've never used LR. So LR does not use the same engine as ACR?

It does use the same engine. I'm referring to Photoshop's Info palette with respect to it's Lab values.
Note too, just because LR and ACR use the same engine, not all is equal. Case in point, the soft proof readout's in LR6 are WRONG. The same values in ACR and LR5 match and are correct.

[BobDavid]

Mr. Rodney, I always chuckle when I see your avatar--A Nipper-esque dog staring into a CRT monitor. Having studied sensation, perception, and the neurology of domestic dogs, a few things come to mind: 1) Dogs are farsighted; 2) Dogs are dichromats. They see muted blues and yellows. The ratio of rods to cones is higher, much higher than it is in humans; 3) Not only do dogs have more rods than humans, they are able to distinguish a wider range of luminance values; 4) Despite being able to discern more shades of grey at night, their visual acuity is not great; 5) Dogs have a visual streak that runs along the X-axis of their retinas. The visual streak is jam-packed with motion receptors. They are able to detect motion that is imperceptible to humans (that's why they are good hunting companions); 6) Dogs process more frames per second than humans. Although the amount of visual information per frame is less for dogs than it is for people; 7) Dogs are able to detect as much, if not more, information through their sense of smell than humans do with their eyes; Dogs are able to hear sounds from four times the distance of a person with normal hearing; 9) Dogs have a wider hearing "gamut" than humans by approximately 2X. 

[digitaldog]

Mr. Rodney, I always chuckle when I see your avatar---Nipper staring into a CRT monitor. Having studied sensation, perception, and the neurology of domestic dogs, a few things come to mind: 1) Dogs are farsighted; 2) Dogs are dichromats. They see muted blues and yellows. The ratio of rods to cones is higher, much higher than it is in humans; 3) Not only do dogs have more cones than humans, they are able to distinguish a wider range of luminance values; 4) Despite being able to discern more shades of grey at night, their visual acuity is not great; 5) Dogs have a visual streak that runs long the X-axis of the retina. The visual streak is jam-packed with motion receptors. They are able to detect motion that is imperceptible to humans (that's why they are good hunting companions); 6) Dogs process more frames per second than humans. Although the amount of visual information per frame is less for dogs than it is for people; 7) Dogs are able to detect as much, if not more, information through their sense of smell than humans do with their eyes; Dogs are able to hear sounds that are four times the distance of a person with normal hearing; 9) Dogs have a wider hearing "gamut" than humans by approximately 2X.

That's why when I speak of the perception of color, it's always human color perception. I'd never speak to the perception of dogs, donkeys or dinosaurs. 
As a dog owner who understands their keen sense of smell, it's huge smell gamut, I have to wonder why they have to get their noises so close to any and all poop they encounter. But that's pretty OT...

[BobDavid]

That's why when I speak of the perception of color, it's always human color perception. I'd never speak to the perception of dogs, donkeys or dinosaurs. 
As a dog owner who understands their keen sense of smell, it's huge smell gamut, I have to wonder why they have to get their noises so close to any and all poop they encounter. But that's pretty OT...

...interesting comment about dogs sniffing other dogs' deposits. I suspect the closer a dog sticks his noses to the target, the more data he acquires. Much like the way I move my head closer to a newspaper to make out the words. OR, use Ctrl + to enlarge a picture on the monitor.

---thanks again for taking the time to explain color management/workflow. It is often a vexing process for laypeople such as myself.

[Simon Garrett]

Back to the topic, if you want to understand the history of colour management see here.

[Schewe]

Hum, I guess I'll need to look into what Lee said and get back to him :~)

I tried listing to his video but I literally fell asleep. Color management isn't very exiting in the first place but Lee's voice lulled my to sleep. Not sure I want to try again :~) But if he said use Adobe RGB in 8 bit/channel, I need to "fix him" some...

[Tony Jay]

I tried listing to his video but I literally fell asleep. Color management isn't very exiting in the first place but Lee's voice lulled my to sleep. Not sure I want to try again :~) But if he said use Adobe RGB in 8 bit/channel, I need to "fix him" some...

Priceless! 

[Tony Jay]

Back to the topic!

I think that any assumption that using a wide gamut colorspace - in this case ProPhotoRGB - somehow leads to garish, inaccurate,or unrealistic colours cannot go uncontested.
The colorspace itself does not cause these problems and the same problems can arise using sRGB.
The way we edit colour is much more likely produce these issues than the editing colorspace per se.
In fact, I want to edit in ProPhotoRGB precisely to avoid many of these issues.
ProPhotoRGB allows me retain many of the subtle colour hues in my prints that would be lost in a smaller editing colorspace.
(Maybe this is what Andrew was referring to when mentioning preserving differences between colours.)
Those subtle colour hues give a realism to a print that would otherwise be unobtainable.
Using a smaller editing colorspace does not preclude subtle colour gradations but if the baseline colour of an area is outside the gamut of a smaller colorspace then the result looks unrealistic (desaturated), to my eye anyway.

Another principle of digital editing workflow that needs to be reinforced is to retain as much data as possible through the process to maximise the output image quality.
Using a wide-gamut editing colorspace is one part of this but the other would most definitely be editing a 16-bit, rather than an 8-bit file.

Also, any suggestion that colour management is immature just because monitors gamuts cannot encompass the ProPhotoRGB colorspace is incorrect. As Andrew has already pointed out some of the plots in the ProPhotoRGB colorspace do not map to colours perceptible to human colour vision. But, even if they did, current, and foreseeable, technology does not allow monitor projection to display all these colours.
But the most important fact to point out is that, by volume, the vast majority of colours in an image will be potentially displayable with current wide-gamut monitors.
I have never seen a colour shift between an image displayed on a monitor versus a print of that image when there are colours represented in print that cannot be displayed by the monitor. (This presupposes a good colour management workflow in the first place). However the print might look more detailed (partly because the printed resolution is better than the monitor resolution) but also because all those subtle hues and colour gradations that were not visible on the monitor area actually translated to the print.
One does not need to "see" every colour potentially in an image and representable in ProPhotoRGB to have a robust colour management workflow.

My comments are valid in the context of the print being the prime output destination.
There is nothing inherently evil in using a smaller colorspace - especially in output - or even an 8-bit file.
This is true even when using the sRGB colorspace for JPEG's created OOC as long as it is understood that at that point the goose is cooked.
Many people's workflow mandates producing 8-bit JPEG's in the sRGB colorspace OOC and that is fine.
If however, one is not forced into a workflow like that then preserving as much image data as possible as late into one's workflow just seems prudent since it contributes to maximising image quality.

Tony Jay

[D Fosse]

ProPhotoRGB allows me retain many of the subtle colour hues in my prints that would be lost in a smaller editing colorspace.

<sigh>

Not to beat a dead horse, but this feels like an uphill climb.

If details are lost, you have clipping. You don't automatically get clipping in any other color space than ProPhoto. That's the myth. Not if you do it right.

I never accept clipping, ever. Just so we're clear on that.