ProPhotoRGB

[digitaldog]

The last time I tried to do some comparison print tests between perceptual and relative colorimetric with an image with out of gamut colors, I couldn't see any differences.

You might want to try this on a synthetically constructed image like a Granger Rainbow (easier but not as effective, a good old spectral gradient).

Advantage; you’ll see every tiny bit of difference on such an image. You can build the image in LAB then convert to sRGB, Adobe RGB (1998), ProPhoto RGB or just assign new working spaces to the image in one of those RGB color spaces (the numbers don’t change so it doesn’t matter). What’s interesting is to see how the source profile affects the final. That is, the numbers in a duplicate file are identical but by assigning sRGB or ProPhoto RGB and converting to the same output profile, you’ll see differences and how the source plays a role.

Disadvantage is this isn’t a real image. However, when you see banding or other issues in such an image, it does indicate it IS possible to see this in a real world image! I recently did a lot of testing of two profile packages using this and real world images. We were trying to see the effect of smoothing on the output profile. On even a good deal of representative test images, it wasn’t always that visible where there were issues with the profiles. IF the image had a certain range and saturation of color somewhere in it, you could make out a slight difference. With the synthetic image, it jumps off the page!

So what does this tell us? Real world images are what we all work with. However, it’s often the case that issues like banding or smoothing issues will not be seen in some images, will be seen slightly in others and look pretty obvious in yet other real world images. It’s kind of a crap shoot. With a synthetic image, there’s no ambiguity what’s going on.

Like the other post in this forum about LAB conversions, there are those that say “you can’t see the effect of this or that.” That is somewhat valid but we never know when we could get bit in the butt based on image content (and the output device). The data loss or issues are still there, so if we can’t see them, does that mean we should ignore the issues or pretend they may not crop up later?

So using synthetic images has some merit just don’t put a huge amount of concern in the results all the time. Ideally, test with both types of images. They both tell you a lot about a process!

[PeterLange]

 
O.k. Mark,

Here we go. Today I started out to capture the famous yellow flower (at sunny daylight); colors even fall out-of-adobeRGB.  In fact, I made two shots: sRGB JPEG and Raw.  The Raw file was processed at camera-default settings to ProPhotoRGB, 16 bit.  The only manipulation involved (by purpose) was that I set all sharpening to zero; in-camera as well as in ACR.

Both files were soft-proofed to an Epson premium glossy profile.  The paper doesn’t really matter, however, the scenario is clearer because the sRGB JPEG file is completely in-gamut now.  Whereas the ProPhotoRGB file shows many many out-of-gamut marks with regard to the printer profile.

Surprisingly, the sRGB JPEG file shows much more details (at 200% magnification), whereas the ProPhotoRGB file shows posterization.  Changing the SoftProof from RelCol to Perceptual helps, but not perfectly.

How can this be? It seems to me that the in-camera conversion & gamut compression to sRGB is somehow realized by a proprietary perceptual algorithm. No / less channel clipping, unlike RelCol. Remember, if the color engine supports this, it’s even possible between matrix spaces (afaik). If someone knows more, please post. Finally, saturation is sacrificed while details are maintained.
Also, it’s obvious for me that the Perceptual intent with the printer profile does not differ enough from RelCol to deal with such remote out-of-gamut colors from ProPhoto RGB.


This does not claim to be a scientific test.  Equipment or whatever may play a role. Others may get different results or see things differently. A round-robin test could help.  Also I desisted from any image editing, which of course requires only basic PS skills to realize a perceptual de-saturation by hand.


Do I like the results – no.
Do I want to go back to sRGB JPEG – no.   
Am I missing something here – can’t exclude!
Do I go on holiday now – yes, soon  .

Cheers! Peter

--

[PeterLange]

Following please find an early article by Bruce Fraser, where he exactly describes the problem which I encountered:

Extracted from:  “Out of Gamut: Exploring Wide, Open (Color) Spaces”
http://www.creativepro.com/story/feature/8582.html
(Bruce Fraser, Sept. 2000)

“A second problem arises with the Perceptual rendering tables built into most output profiles. All Perceptual tables contain an assumption about the source space: Some profiling tools, notably Pictographics' ColorSynergy, actually allow you to choose a source space, but most use a hardwired assumption. Few if any use a space anywhere near as large as either EktaSpace or ProPhotoRGB, so the default gamut mappings may not work all that well.

You can get around this -- very successfully, I may add -- by working inside a simulation of your output space and adjusting the colors in the source image to achieve the appearance you want. Photoshop 6.0, which should ship in the next few weeks, offers better output simulations than any previous version, and lets you preview RGB output as well as CMYK. So you can regard the gamut mapping issue as a blessing and a curse. The blessing is that you can control the mapping of the colors in the original to your output. The curse is that you have to.”


For further theory just follow my posts from page 7.  Finally, I just can recommend to try the real-world tests which I suggested at page 10:  Capture an intensive yellow flower at sunny daylight.  Make two shots: sRGB JPEG and Raw.  Process the Raw file in the way you prefer to ProPhotoRGB.  Softproof both files to your preferred printer profile (or make tests prints) and compare details.

Peter

--

[Ray]

Well, that's killed the thread, hasn't it! We're all in agreement   .

[digitaldog]

The only con to ProPhoto is a slightly greater chance of posterization or banding if editing in 8-bit mode. But if you use 16-bit mode (which you should be doing anyway) this is irrelevant and there is no real downside to ProPhoto.

The “downside” is a heck of a lot of potential colors you can’t see. You might be editing colors out of display gamut that fall within working space gamut. For 99% of users out there, this is the case with Adobe RGB (1998) as ProPhoto RGB. Unless you have one of the new Adobe RGB (1998) gamut sized display that is. This is a minor issue but one users should be aware of. I’d rather have the colors at my disposal and be very careful about editing extreme saturated colors (IF they happen to be in the image) then not have them to begin with.

[Ray]

As for the idea of a wider space only being useful for very saturated colors, there’s more to it than that.

My non-technical understanding would be, if I have a pure, saturated yellow paint in my pallette, then I can mix it with other colors to produce shades that I couldn't get if my yellowest yellow was not pure, say slightly green or blue.

I've just tried a little experiment setting 'proof setup' to a Bill Atkinson Premium Gloss profile for the Epson 9600. My eyes have already told me I will get a purer yellow on this paper using ProPhoto, so I was curious to see what the numbers are.

For pure RGB yellow (255,255,0) ARGB is within gamut for this paper at maximum saturation, ie. it's not possible to make it out of gamut. However, using ProPhoto the saturation has to be brought back a small degree (-6) with the hue/sat control in PS. The proof RGB numbers, second reading on the info pallette, are 254,255,28 for ProPhoto and 242,255,81 for ARGB. The numbers tell me the ARGB yellow at maximum saturation has more blue and green, and that's what I see on the paper.

However, all the other primary colors at maximum saturation are out of gamut by a significant degree for both color spaces in relation to this glossy paper, particularly green and cyan which have to be brought back by 62 points on the hue/sat control in the case of ProPhoto and 47 and 29 points respectively in the case of ARGB.

[digitaldog]

First, with rendering intents to output spaces, you have to decide if you’ll clip the colors or compress them. ...

Just a friendly question:

Are there any printer profiles
which support Perceptual rendering in way
which assumes that the source space was ProPhotoRGB?

Peter

--

No. Output profiles don’t have any idea when built what the source color space will be.

[Mark D Segal]

Ray, my understanding of how all that you just mentioned works, in a nutshell, is as follows. D50 and D65 are two conventions for the color temperature of the monitor, and the color management folks tell us if we use 6500K on the monitor in our minds eye the image looks closer to a daylight reflected image than it would if we used 5000K monitor temperature, eventhough daylight lamps are in the range of 4800-5000K. When you ask them why, they just tell you it is observed experience about how we humans function (whoever said coherence and consistency were hallmarks of the human condition?).

The main parameters for calibrating a monitor are the white point, black point and gamma. Once these are set, profiling is done to measure the numerical differences between the (correct) data describing the color reference patches of the profiling software and the (perhaps incorrect) values the monitor shows them as. These measurements are then used to create a profile that neutralizes these differences so the monitor shows the colors of the file data. Likewise the printer profile measures and neutralizes differences between the (correct) data in the software file for each patch on the profiling target and the (perhaps incorrect) numerical values the printer prints them at. (Both sets of profiles calculate interpolations for all the data between these reference points.)

The file data is "nerve central". Once these two profiles are in place, the system is color managed because both the monitor and the printer are being managed by the two profiles to reproduce the same set of numbers in the file. All file data that is in gamut for both the monitor and the printer/paper should look pretty much the same, allowing for the fact that transmitted and reflected light are inherently different. We cannot see on the monitor colors that are out of the monitor's gamut. As there is no reason for the monitor gamut and the printer/paper gamut to be the same, Photoshop's soft-proofing module simulates an impression on the monitor of what the printed image will look like once we activate it with the printer profile that will be used for printing. Note, we are not profiling the monitor with the paper profile. Photoshop just uses the paper profile to create the soft-proof.

As far as I have been able to parse it, this whole discussion about whether or not it is useful to work in ProPhoto color space so far boils down to two propositions: (a) if Prophoto includes - but ARGB98 excludes - values that current and future generations of printers can reproduce, we should retain those values because they will improve printed image quality; ( as long as we work with 16 bit data, there is very little risk that Photoshop's shakedown of out of (printer) gamut colors will produce any noticeable banding.

[paulbk]

Aside: “using Soft Proofing correctly, you can be about 90% accurate in your visual predictions on-screen to print.”
As I said, using a large color space puts the fun and intrigue back in printing.

Seen from a different point of view: If you send the same file to the printer you get the same print every time. Even if half the pixels are out of print space gamut. The print process is deterministic going one way. But the reverse is not true. If you could run time backwards and feed the same print into the printer you will not get the same file every time. Because there is only one in gamut file and many many out of gamut files that will render to the same print.
True?

[PeterLange]

Peter, I don't think that is correct. Firstly we are not converting a color space to a printer profile. As I mentioned above the printer profile measures and corrects differences between the file data and how the printer reproduces that data.

Mark,

I’m not sure if I get your point.

RelCol is always RelCol (afaik).  All in-gamut colors remain unchanged, whereas all out-of-gamut colors are mapped to the surface of the target space.  The more you have, the more it leads to an overpopulation = posterization = loss of details (also called clipping which may be a little bit misleading).

Peter

--

[Mark D Segal]

My thinking exactly. Some of this stuff can be in the realm of angels dancing on the head of a pin; that said, it's still fun being just geeky enough to get under the hood and better understand some basic principles at play.

By the way, about a year ago I also tried alternative rendering intents on a photo of some buildings and trees taken in downtown Grenoble, France. Using my Epson 4000, Enhanced Matte paper and Epson's profile, I DID see subtle differences in some of the shades of green, sky hue and colours of the walls - but I emhasize the differences were subtle and most noticeable with Absolute Colorimetric; this was to be expected (Real World Color Management, page 92).

[Mark D Segal]

Andrew, what I take from your nuanced response is that there is a (perhaps very) low probability risk of encountering some "issues" (whatever those are) in normal photographs using ProPhoto colour space, but - going back to earlier posts - perhaps this risk is worth taking, because we know that some of today's printers can reproduce some colour values that ProPhoto retains but ARGB98 does not and this is likely to become more pronounced with future improvement in printer technology. Would this be a reasonable way of drawing together the key implications of this whole discussion?

[Ray]

Insure the output profile is accurate, especially with respect to the soft proof table (the table in the profile that controls the soft proof, not the actual output)

Andrew,
I don't have the equipment to make my own profiles. The profiles I used with my Epson 1290 are Pantone created profiles freely available for various Epson papers and recommended as being quite satisfactory by Ian Lyons.

Now I know they are not going to be the best. There's variation amongst desktop printers, just as there is with lenses of the same model. But I found these profiles to be good enough at the time. I now use an Epson 7600 for critical work.

The example I've been using to make a point about the gamut warning function is probably less relevant with modern papers, inks and profiles which probably all have fairly close gamuts. For example, with a saturated image open on screen, gamut and proof colors ticked, black point compensation and 'simulate paper white' ticked in custom proof setup, one can switch between the various profiles that are in one's color folder and observe the effect any particular profile has on those ugly gray patches.

For example, I've got an image in PS at the moment I've been workin on. I've increased the saturation a lot because I like saturated colors. With Epson's 7600 Watercolor profile (device to simulate) there's a moderate degree of gray patches; with Textured Fine Art paper there's a very slight degree of gray patches; with Premium Lustre there are no gray patches whatsoever. However, with the Pantone/Epson PhotoPaper profile, virtually half the image is covered in gray patches. Photopaper is an old paper type. It doesn't appear to have nearly as wide a gamut as Epson Premium Gloss and from past experience, if I were to print this image on Photopaper and let the rendering intent sort things out, then shadow detail would be severly compromised in a way that is not visible on my calibrated monitor in proof setup.

I've assumed all along that the gamut warning is giving me a fairly accurate indication of the comparative size of the gamut of any particular paper type. But if you're saying this gamut warning in PS is basically a useless piece of crap, then I'm natuarally wondering what tool could replace it   .

[digitaldog]

“..if your original capture does not contain tones out of the gamut of AdobeRGB, then there is no point whatsoever (and quite a few negatives) in using ProPhoto as your color space unless you plan to artificially manipulate the saturation of your image to a considerable degree.”

I’ll go one step further, and I know nothing about color management. You don’t have to be a color scientist to understand this: If you are working in a color space you can’t print, you are doing a blind edit. That’s a fact on it’s face. The rest is a matter of degree.

What’s amusing are those who say they are preserving colors by working ProPhoto even though they know they can’t print the colors they are “preserving.” And worse, they can’t even see some of the colors they are preserving. It boggles.

Working in 16 bit is an entirely separate issue.

If you’re shooting RAW to some color space on the fly, you can’t know what scene may or may not contain data beyond Adobe RGB (1998). IF you’re shooting RAW, you can select, as I do when viewing the ACR histogram. No harm done.

As for the gamut of the printer, what printer and when? If you look at the gamut of printers in the last 10 years (a century in dog and digital years), a lot has changed. I have no idea the what I may be able to print with a 12 color ink jet or some other technology in 5 years or less. As such, I find no reason why I should funnel color I can capture into a smaller space based on what I may be able to reproduce today. If 100% of my work went out to a four color press, this would be a moot point. I’ve got right now, a Pictrography 4500, Epson 2400, 2200, 980 and R800. I have no idea what I’ll be using next year, let alone 10. There’s no harm at all in using a wider gamut container (editing space) in 16-bit considering my crystal ball isn’t always working with respect to future output technology.

There’s no output device that can print sRGB other than a display. So the argument that there’s no device that can output ProPhoto RGB isn’t compelling. Working space are not output spaces and vise versa.

16-bit isn’t an entirely separate issue since the bottom line here is what data can you capture and keep and what should you keep?

[paulbk]

There is nothing said thus far to refute the premise of Jeremy Daalder's article. This is not about opinion. It’s about numbers. Predictable, deterministic math. It all boils down to a squirt of ink determined by a number. It’s as simple as that.

I’m not saying don’t use PhoPhoto. Use it. Have a ball. I am saying that when your working space is far greater than your printer/paper space you are knowingly introducing an unnecessary unknown. Which adds mystery to the output but nothing else of value.

If some one can explain how a color outside the print space is of some value during editing, I’ll read it. But at the moment, it’s mere religion.

[Ray]

I see no one has answered my question regarding the actual average range of discrete colors in a 24 bit 1Ds image. I thought it might help to get in perspective concerns that some have about ProPhoto and 8 bit. There are programs for calculating the actual numbers of different colors in any given image. I vaguely recall someone with the British Journal of Photography once used such a program on a selection of images and found the numbers surprisingly small, like 500 to a few thousand.

Despite using ProPhoto myself for all the reasons outlined in this thread, I've never done any rigorous comparison of a selection of real world images, editied in both ARGB and ProPhoto before printing. Ultimately, the proof of the pudding is in the eating. I suppose I wouldn't expect to see any significant difference in the results unless I knew precisely where to look, or how to select  images that could reveal the differences.

For example, if I wanted to demonstrate the differences between sRGB and ARGB, I would choose an image with lots of solid green, the green of leguminous plants and freshly fertilised lawns, and lots of subtle shades of cyan. I would expect, no matter how hard I tried, I would not be able to get a green, green enough with sRGB (too much yellow), and some of the subtle and distinct shades of cyan would merge, compared with the ARGB rendition.

To demonstrate the differences between ProPhoto and ARGB, the obvious choice of image would be one with bright yellow flowers as well as canaries, but also subtle shades of dark brown, (and what else? I'm not sure.) It should be possible to produce an image containing a series of dark tones that are distinct when printed from the ProPhoto color space, but not distinct (merged) when printed from ARGB.

[Schewe]

In workflow A camera colors are clipped in 2 steps, in workflow B in one step.

Exactly. . .in 'A' you clip the color before even bringing it into Photoshop-thereby forever giving up the possiblitity of using the color data that get's clipped. In 'B' YOU have control over what does and doesn't get clipped and how it gets clipped...so, the question is, do you want to have control? Or, do you want to toss camera colors before you even get the image into Photoshop?

Me? I'm a control freak-I want total control. I don't want color that the camera can capture to be clipped merely because the working space can't contain it.

sRGB clips a lot...
Adobe RGB clips a little...
Pro Photo RGB doesn't clip...

What you do with your image is up to you, but personally, I don't want to lose anything. That's why I use 16 bit and Pro Photo RGB.

Don't Fence Me In...pretty much says it all. Adobe RGB is a fence that reduces what the camera can capture. I think the grass is always greener on the other side-and in the case of Pro Photo RGB, that is exactly the case!

:~)

[PeterLange]

Me? I'm a control freak-I want total control. I don't want color that the camera can capture to be clipped merely because the working space can't contain it.

Jeff, thanks a lot for explaining the philosophy.

Peter

P.S.: I have no luck with the Color Range “Out Of Gamut” option.  Selection always differs from the SoftProof / out-of-gamut marks.

--

[digitaldog]

The Granger Rainbow is certainly brighter and more colourful when assigned a ProPhoto profile (as opposed to sRGB and ARGB) both before and after proof setup for Premium Glossy, RelCol, simulate paper white and  sat reduction to tame gamut.

Brighter appearing when you Assign or after you’ve output the file?

Brighter appearing due to the new meaning of the numbers certainly makes sence. But the numbers haven’t changed (only the definition). This is where this exercise is interesting. Let’s distill the Granger Rainbow down to a single pixel, say R255/G0/B0. You’ve built this one pixel (and in fact the entire Granger Rainbow) but it has no specific color space associated until you assign a profile. If you assign sRGB, the number is still R255. If you assign ProPhoto RGB, well the number is still R255. The only difference is you tell the color management system the scale of the R255.

So we take three identical files, all filled with R255 but define where in the color space that most saturated red lives within context of human vision. That info is used when we convert to an output color space. The new number will be different for each of the three files (sRGB, Adobe RGB (1998) and ProPhoto). We’ve used the same output profile, the same rendering intent and the same original R255 value. This little tests is useful in showing the effect of the source definition of the color space. It’s interesting when using the same output profile for the same device but built by differing packages (especially with Perceptual rendering).

[digitaldog]

The Raw file was processed at camera-default settings to ProPhotoRGB, 16 bit.

And that alone could be a major factor in how well or poor your results are with no bearing on any color space you converted the file into.