Is ProPhoto and colour management in general worth it?

[Garbz]

I am starting this thread to open up a full discussion on colour management, based on the recent threads asking about profiles, and other questions I’ve received. I intend to give a bit of info here from what I’ve learned the past 6 months since taking on a fully colour managed approach to photography. But I mean this to be a starting point. There may very well be some misinformation in here so feel free to point out anything.

What I present here is intended to be simply the what and why of my approach to NOT using ProPhoto and rarely even using AdobeRGB for my photos. For those who want to skip to the end there’s a summary in the conclusion. It covers some basics and some advanced themes at the same time. So let’s get into it.

Intro to ProPhoto:
For those who haven’t heard of colour manangement or gamuts the concept is simple. The reddest red in your photos (255,0,0) is not the reddest red in existence. It’s simply the reddest value your screen displays and for most of the screens out there which use the industry standard colour space sRGB, this isn’t very red at all. The gamut is a diagram showing the colours. The CIE1931 diagram below shows the 3 gamuts talked about here. The smallest is sRGB, the middle is AdobeRGB, and the largest is ProPhoto.

The good parts about this is obvious. ProPhoto can display a far wider range of the visible values than any other space meaning it can cover nearly all colours that exist and is far larger than any device can actually currently display. Right now down to the bad.

Bit Depth:
The first problem that many don’t realise is part of the principle of wider colour gamuts. There are more colours to represent. There are weekly discussions on this very forum about the benefit of using 16bit processing with the existing sRGB gamut to improve processing quality. The use of ProPhoto strains this further. An example using the Relative Colorimetric rendering intent, arguably the best for photos:
sRGB(255,0,0) = ProPhoto(179,71,27)
sRGB (0,255,0) = ProPhoto(138,237,78)
sRGB(0,0,255) = ProPhoto(86,35,235)

So to display the same colour in ProPhoto uses less of the 8 bit space (which is obvious from the gamut above). Now assuming you have all the extra information to make good use of the ProPhoto space this is fine, but for content that may not contain colours that rich to begin with you are just wasting valuable bits which could be used to more accurately represent the colours you have.
Fortunately the fix is easy. Work in 16bit mode at all times when using wider gamuts. Using ProPhoto for an 8 bit image is taking a step back in quality not forward.

Real Colours:
Still talking about bit depth ProPhoto(0,255,0) and ProPhoto(0,0,255) don’t actually exist. They are off the spectrum of colours present and exist only as a mathematical tool to help convert lesser colours. This is even more of a reason to use 16bits to edit such files as bits are wasted on imaginary colours. Admittedly this is minor, just pointing it out.

Processing Displaying and Printing:
This is the one that gets me the most. For people who do colour critical work why work on something you can’t see. Few screens display the AdobeRGB gamut even, NO screens display the ProPhoto gamut. HPs new colossal screen comes somewhat close but I don’t think anyone here will sell their car to buy one. I am willing to bet most of the people here have a TN panel LCD in which case they can’t even produce the sRGB space with consistent accuracy since they have a 6 bit output.
Even if this is not the final output intent (if it’s for screen you really better be using sRGB since I guarantee that that’s what 99.999% of the world will see it in), even if you will produce some nice wide gamut prints it makes little sense working with colours you can’t see since the output becomes un predictable.

The point is with a wide gamut screen the AdobeRGB space makes sense since this is a standard many wide gamut screens use, but going further is again one step towards pointless. For output the same thing happens. You need a very fancy printer to get this colour purity out on paper. Often the paper itself becomes a limiting factor. I’ve used the finest lab in Brisbane I can find a few times (this doesn’t really say much but bear with me) and they have always asked for AdobeRGB files because their fancy printer doesn’t print much outside of this colour space anyway. A few other labs use this profile too and it somewhat seems to have cemented itself as a standard for colour critical work. ProPhoto has not.

The Source:
With the output sorted meet our model for the day:

For those using a colour managed browser the photo on the left is in ProPhoto space. For those who don’t, download it and open it in windows picture and fax viewer. (See how much hassle this colour management thing is, but that’s not the point.) More importantly do you see a difference?

Now our blood thirsty model above was opened from the RAW in ProPhoto space in 16bits. And then the image was artificially saturated. Yes even this image taken on a bright sunny day was saturated further and is more colourful than a real image. I chose this image because red and green are colours that make a difference. Blues are very pure across all colour spaces.

Now it is widely known that cameras capture larger gamuts than the sRGB space. So with that in mind lets check exactly how much more we get out of the world.

The image on the left is the ProPhoto original, soft-proofed to sRGB, and the image on the right is soft-proofed to AdobeRGB. The grey pixels indicate out of gamut colours. So here in our artificially saturated scene a very tiny percentage of the image is actually larger than the sRGB gamut. Nearly none of it is outside the AdobeRGB, but it gets even better.

This only shows the out of gamut colours. It does not show what the actual image would look like after conversion. The process of colour space conversion is somewhat more complex than that. It will do subtle actions that aren’t even notable to me on a wide gamut monitor. Depending on how it’s set it will either shift the out of gamut colours to the nearest reproducible colour, or slightly adjust the saturation of the entire image to keep the relationships the same and move the colours all into the visible.

In this case the relative colorimetric rendering intent produces an sRGB image that is entirely visually indistinguishable from the ProPhoto to the naked eye, even on a wide gamut monitor. To show exactly the differences below is a difference taken of the sRGB layer converted back to ProPhoto (remember it’s lossy so the new old pure colours are not recovered). Now considering just how hard to see this is I adjusted the levels of the result quite severely the settings: (0, gamma 1, 40) and displayed it on the right.

I haven’t done the math but I’m willing to bed that the deltaE of the different colours is so low that most people could barely distinguish them side by side let alone be accurately reproduced during printing. Which brings me to my final point.

Colour conversion is not an exact science:
To start with the conversion is lossy. Like going from 16bits to 8bits you throw away the colour information. But worse yet for each set of colour profiles there are different methods of converting from the profile to a standard conversion space and arriving at the new profile. Relative Colorimetric, Perceptual, Absolute, and saturation adjustment. In each case there are settings such as black point compensation, and dithering. Furthermore there are various processing engines. The windows ICM engine and the Adobe ACE engine are the ones used by Photoshop. And now unless your process your own colours on your fancy expensive screen, and print them on your fancy expensive printer there is no way of knowing how the conversion will be handled. My lab didn’t tell me, and neither did the guy who looked at my webpage in a colour managed browser. Does it make a difference? Here’s our blood sucker again converted from ProPhoto to sRGB using Absolute colorimetric conversion, with dithering enabled. On the left via AdobeACE, on the right via Windows ICM.

Wow there's a character limit. Continued below:

 

[Garbz]

So now even soft-proofing fails if you don’t know EXACTLY how the printer will work. In Photoshop you can take control and kill colour management in your print driver, BUT without expensive printer calibration tools (more expensive than for a screen by far) you are relying on the profiles made by the manufacturer, something which changes depending on settings and paper. And if you print in a professional lab you’re relying on their good judgement only.

Conclusion:
So does that make ProPhoto useless? Not really. If you are photographing some wicked artificial lighting in a nightclub then maybe it will be of some benefit. And if you control the colour process from start to end it will be of some benefit. AND if you can actually display the images or print them with a wide gamut and make your process more than an exercise of academia , turning it into something useful there will be some benefit.

The summary I promised as to why I don’t touch it:
- I don’t wish to waste the bit depth on something that yields little benefit to me.
- I can neither display nor print the images.
- As such I cannot reliably process the images (less of an issue compared to the others)
- I rarely have a use for it given my most saturated images rarely register a difference. (Sunsets are an exception here)
- Colour management is approximate voodoo, and complicated to boot. Ignoring the issue in this case is for me a perfectly viable alternative.

Phew. Well what do you think?

 

[Alpha]

For me, the easiest way around this senseless problem is, I feel, relatively intuitive. Edit and file-share in Lab color, which is an identical gamut on everyone's machine no matter what their monitor says. Soft-proof when possible. And only print on printers or with printing companies where you know the setup is calibrated...so apparent color differences that result from conversions to RGB or CMYK color spaces really become negligible in terms of workflow. Be fastidious, pay attention, work with people you trust whose equipment you know, and do what you can to make your life easier. I think it's really that simple.

As you know, Lab is technically a wider gamut than even ProPhoto, albeit in a very different way.

 

[Bifurcator]

I'm still reading this but here's something for you while I'm going over it:
http://www.imaging-resource.com/NEWS/1213841085.html
http://www.tomsguide.com/us/dreamcolor-hewlett-packard-monitor,review-1096.html

 

[Garbz]

Quite technically LAB colour is the maths behind colour conversions. When you convert ProPhoto to sRGB it uses a matrix transform to get to lab first with a fixed rendering intent, and then uses a user selectable intent to get to the final space. So yes LAB is the guts behind it all and it is good to edit it, but it does not solve the problems of colour management.

Specifically, monitors don't display the LAB space so you still need to convert to RGB first (photoshop does this for you behind the scenes) and at the same time the screen does not show the entire LAB gamut. Well technically nothing does. It suffers the same bitdepth problem as ProPhoto because of it's theoretically visibly infinite gamut so 16bit processing is a must (again this isn't a problem really). But in the end how do you get to the printer. Yes only use a lab with proper calibration but the LAB -> RGB process, or LAB -> CMYK process still has use selectable rendering intents.

While we hope that the people know what they are doing. The point of this was that colour is far from an exact science and the benefits of ProPhoto are often outweighed by the other varying factors in many people's photographic method.


Bifurcator why didn't you link that in the monitor thread. That screen still only has 133% NTSC coverage so ProPhoto still has a wider gamut. Still the problem of getting the photos out is still prevalent. It's all good and fine if I can see the picture in it's full glory but how do you share it with the world? Lets hope that in 10 years we'll all have this screen and this thread becomes redundant

 

[Bifurcator]

Bifurcator why didn't you link that in the monitor thread. That screen still only has 133% NTSC coverage so ProPhoto still has a wider gamut. Still the problem of getting the photos out is still prevalent. It's all good and fine if I can see the picture in it's full glory but how do you share it with the world? Lets hope that in 10 years we'll all have this screen and this thread becomes redundant

I didn't know about it. In reading your thread (quite interesting) I just got the urge to google for "10 bit LCD" and there it was - released 3 months ago and for a reasonable price (relatively speaking).

I think in 10 years most of us won't be here with all the active large-scale eugenics going on and the revitalization of the nuclear threat of holocaust - but yeah the logic is right. I mean we all used to be on green screens so the first color monitor users must have felt the same way in many respects. Still even without being able to share the color images many people bought the color monitors.

Yeah it's basically just slightly bigger than Adobe RGB but not yet Pro Photo. Here's it's spacial diagram plotted in 2D:

The legend is actually wrong with semi-transparent fill colors used instead of area border colors but you can see the yellow triangle representing the monitor clearly overlaps all other spaces with some NTSC colors escaping out the blue/green side. The Adobe RGB triangle is the second largest next to the monitor itself with the monitor being stronger in the red/blue areas which actually brings it pretty close (or at least closer) to Pro Photo I think.

 

[Bifurcator]

Even if this is not the final output intent (if it’s for screen you really better be using sRGB since I guarantee that that’s what 99.999% of the world will see it in), even if you will produce some nice wide gamut prints it makes little sense working with colours you can’t see since the output becomes un predictable.

I guess it's actually 70% ~ 80%. People using Safari or FireFox 3.x will see a difference if there is a difference in the image at different color space assignments. And the monitor while stepped and dithered even at 6 bits shows these differences.

When you say the output is unpredictable what exactly do you mean?

You need a very fancy printer to get this colour purity out on paper. Often the paper itself becomes a limiting factor.

With all the dithering that standard ink-jets do I would think that this is true. They don't print solid colors at all you know. Of course there is a massively HUGE difference between photo grade paper made for the printer by that printer's manufacturer and standard copy paper. For Canon there's

Photo Paper Pro II,
Photo Paper Pro,
Photo Paper Pro Platinum,
Photo Paper Plus Glossy II,
Photo Paper Plus Glossy,
Photo Paper Plus Semi-gloss,
Photo Paper Glossy,
Matte Photo Paper,
Photo Paper Plus Glossy Double Sided,
Fine Art Premium Matte,
Fine Art Photo Rag,
Fine Art Museum Etching,
High Resolution Paper

plus some other misc. forms.

The Source:
With the output sorted meet our model for the day:

For those using a colour managed browser the photo on the left is in ProPhoto space. For those who don’t, download it and open it in windows picture and fax viewer. (See how much hassle this colour management thing is, but that’s not the point.) More importantly do you see a difference?

I see a huge difference yes. The left one almost looks like a tone-mapped image. The one on the right looks pale and over-exposed. Also on my monitor, the left one seems to have clipped the details out of the background people's hair and darkened the trees considerably. I'm using Safari so I didn't have to DL anything to see these differences.

Now considering just how hard to see this is I adjusted the levels of the result quite severely the settings: (0, gamma 1, 40) and displayed it on the right.

I haven’t done the math but I’m willing to bed that the deltaE of the different colours is so low that most people could barely distinguish them side by side let alone be accurately reproduced during printing.

I see lots and lots of differences between the two here. I guess about 70% coverage. You don't see it? It's very very distinguishable.

My lab didn’t tell me, and neither did the guy who looked at my webpage in a colour managed browser. Does it make a difference?

Again I see massive differences. The one on the right has the yellows cranked WAY up. Reds are lighter and less saturated.

 

[Garbz]

Well the bottom set I am showing the differences between an image converted to sRGB using the SAME rendering intent with two different processing engines. This ties into the output becomes unpredictable if you are not in full control of your processing from start to the print, since you are unlikely to be the one at the pro lab that loads the image and sets the printer settings.

As an after thought this may also not be an issue. If the lab you are using has native AdobeRGB support right in the printer driver then giving them an AdobeRGB file should *SHOULD* mean no change.

The black one yes there is a difference, the left one was so dark that I boosted the brightness so you could see the actual differences more clearly. They are the same image otherwise.

What really perplexes me is that you get a difference in the top one since colour management should make these both identical! This is what I see in both firefox and internet explorer. Firefox makes both images appear identical since they are both converted to my monitor profile. i.e. makes the ProPhoto one look pale, and the sRGB one look saturated (wide gamut monitor remember)

Head over http://www.color.org/version4html.xalter here. This site at the ICC has a test matrix for browser colour support.

 

[Joseph Westrupp]

So, to get to a final 8 bit output, would a good workflow be:

RAW ==> Edit 16 bit PSD in Adobe 98 ==> Flatten ==> Convert to sRGB ==> Convert to 8 bit?

 

[Garbz]

Yes.

Just to point out why for the people who didn't come here after reading another thread. 8bits cover every possible value in the sRGB gamut, but not in any other gamut, so this conversion should happen AFTER the gamut conversion. Also Photoshop processes layers in the currently selected bit depth. So an 8bit image with layers doesn't just have 8bits per pixel, but all internal processing occurs within these values too regardless if you added or played with the layers back when Photoshop was set to 16bit. If you don't flatten the layers before conversion, you are unnecessarily limiting the ability for accurate calculations between layers to be performed.

 

[Joseph Westrupp]

Cheers for the reply. Would degradation be more likely to occur, or would there actually be an advantage, with the substitution of ProPhoto for Adobe 98 in the above workflow?

 

[Garbz]

16bits is a lot of discrete numbers. There's no real downside to substituting ProPhotoRGB in the above, but there's almost next to no benefit either. There's only a handful of monitors with a gamut wider than AdobeRGB (not taking into account my phone which has an OLED screen and therefor an almost perfect gamut), and there's only a small difference in gamut between the top printing money can buy and AdobeRGB. But by all means, there's no field relevant downsides.

Take a look at this, the dotted line is Adobe RGB, and the wavy on is an OCE Lightjet and Kodak Ultra Endura HD (colour laser on chemical paper process which represents about as good as it can get) :

 

[Joseph Westrupp]

Very interesting. There's a hell of a lot to know once you dig a bit into photography.

I presume you have the Galaxy S?

Computer monitors with the pixel density of some of the latest phones would be useful.

 

[Garbz]

Yes I got a Galaxy S. I posted an example calibration: http://www.thephotoforum.com/forum/photographic-discussions/233651-perfect-colour-my-phone.html Interestingly this would come close to outperforming the best monitors on the market today, and it's a technology still in it's infancy

I disagree about the computer pixel density issue for a few reasons. Lack of scalable interface graphics means that there's no option other than tiny eyestrain-o-vision for much of the interface. My 26" screen runs at 1920x1200 and from my seat I can barely make out individual pixels without bending forward as it is. And the biggest killer is fill rate of graphics cards becomes a major hassle as resolution goes up; I don't want to have to buy a $450 video card just to do 2D graphics work. Lightroom is slow enough on this screen as it is.

 

[Joseph Westrupp]

Yes I got a Galaxy S. I posted an example calibration: http://www.thephotoforum.com/forum/photographic-discussions/233651-perfect-colour-my-phone.html Interestingly this would come close to outperforming the best monitors on the market today, and it's a technology still in it's infancy

I disagree about the computer pixel density issue for a few reasons. Lack of scalable interface graphics means that there's no option other than tiny eyestrain-o-vision for much of the interface. My 26" screen runs at 1920x1200 and from my seat I can barely make out individual pixels without bending forward as it is. And the biggest killer is fill rate of graphics cards becomes a major hassle as resolution goes up; I don't want to have to buy a $450 video card just to do 2D graphics work. Lightroom is slow enough on this screen as it is.

Good point; graphics processing requirements would be steeper. I'm just thinking for soft-proofing it would be nice to have more than 70 (ish) PPI for evaluating images, so that they could be closer to final print resolution on screen.

By the way, completely off topic, but do you use DNG?