Why use ProPhoto RGB?

A very nice demonstration, Royce. I will most often use Prophoto RGB for my raw workflow; many, many times I've had the highlight warning in ACR showing clipped reds and yellows of birds, flowers, etc when the conversion is set to output to Adobe RGB, and when switching to Prophoto RGB for the output color space the highlight warning disappeared. Of course, when in Photoshop I will do some curves work, or some local burning of the highlights, to make sure they don't get clipped when converting the color space for output.

I now use Prophoto as my default output color space in ACR (or C1 LE when I'm using it). If I'm working on an image which doesn't contain a high dynamic range I will sometimes switch the output color space of the converter to Adobe RGB since I'm not needing the wide gamut, and even though I'm editing in 16 bit color depth I figure anything that helps smooth tonal transitions is a plus.

I had read the Reichmann article mentioned above a while ago and the match between the 20D color space and Prophoto was a big eye-opener for me. Frankly, I don't see any downsides to Prophoto -- it's another available tool, I use it when it's called for and leave it in the toolbox when another tool is more appropriate for the job. It's all about understanding what the tool is good for and deciding when it's the right one for the task at hand.

For images with low gamut needs, ProPhoto RGB can be a problem. Since it has a larger overall gamut, but is limited to the total number of discrete tones by 16 bit (same total number as Adobe RGB 1998), by definition the distance between 2 adjacent tones must be greater than with a lower gamut color space. So using ProPhoto with a low gamut image can create problems with smooth tonal gradations that would not exist (or would be less evident) on a lower gamut color space.

Hi Greg, can you please demonstrate this with an example?

I understand the theory but would like to see an example of where this "smooth tonal gradation" (or lack thereof) issue can occur in practice.

Thanks,
Eric

Actually I do not since I have avoided using ProPhoto RGB for low gamut images for a few years.

Like Eric, I understand the theoretical reason for avoiding a wide gamut color space for use with a narrow gamut image. Part of the theory is not just that the larger gamut means wider spacing between discrete color tones, but also that some of the ProPhoto RGB colors are purely artificial and do not represent real colors that could be seen, captured or printed. (This is called coding efficiency -- how much of the color space encodes usable colors.) And I trust that a guy like Joseph Holmes has created a highly tuned series of color spaces of varying size gamuts for a good reason.

Having said that, I've been using ProPhoto RGB exclusively for some time now and I do some fairly heavy post-processing. So far have not once encountered a tonal banding problem in a 16-bit workflow. What I have seen frequently is gamut clipping with Adobe RGB, effecting colors captured by my cameras, and printable by my printers. ProPhoto RGB lets me preserve those colors at the cost of being aware of how to work effectively with the color space.

If excessive gamut size was a problem for narrow gamut images I'd think it would create trouble given that applications like ACR, Lightroom and Lightzone have all adopted (variants of) ProPhoto RGB as their internal working color spaces. They do not give the user the option to select anything different the way Photoshop does. Of course these applications are all 16-bit internally as well.

My conclusion has been that gamut size issues are only likely to show up in an 8-bit workflow. Since I never go to 8-bit until the very end of my post-processing stages for output to print or web, my belief so far is that ProPhoto RGB does not have a downside in a workflow designed for image fidelity. The only real downside I've hit so far is not being able to fully see the gamut on current displays, but there are ways to work with that...

Dan Margulis makes the following statements in the 2007 edition of Professional Photoshop: P. 287

"When theory is taken to mindless extremes, we see stuff like the final two alternatives (color spaces). First the practice of using an ultra-wide RGB (such as ProPhoto RGB)... The second is the use of an ultra-low gamma."
Although Margulis sees no harm in using ProPhoto RGB to store the data captured by the camera, he feels that it is presently not useful for color correction.

"Such spaces are so volatile as to make precise moves difficult to impossible...You will discover that it's harder to avoid casts... These spaces are as big as LAB but harder to control."

"The common excuse for these spaces is that it is possible to find colors that only they are capable of reproducing, and that it may be possible to reproduce in print in a few years. If somebody says that to you, I would ask to see an example of the sort of image that he is referring to. If one exists,which I doubt, then I would ask whether it it looks like a rare enough commodity that it could just be handled in its own private RGB when the time came, rather than adopting a ham-handed ultra-wide definition as my permanent RGB working space."

The words in parentheses are mine, but Dan specifically mentions ProPhoto RGB several times during this discussion.
I know Dan Margulis doesn't mince words when he has an opinion, but his techniques have helped improve my Photoshop skills better than any other author, so I'm not in any hurry to change RGB working spaces.

My thought process on the topic has been that I'd rather work with ProPhoto RGB and deal with the minor inconveniences it can cause but be ready when printers and monitors evolve to cover that gamut, than have to go back and reprocess hundreds or thousands of images to take advantage of the new products when they arrive. Once your work flow is set up for ProPhoto and you understand what to watch for, it's not any more difficult to use that Adobe RGB or sRGB. Why stay in those spaces when we know they do clip useful data?
So good points on both sides, but for the types of folks in these forums, I don't see much benefit to waiting to go to a 16 bit space. And I'm no where near as experienced as some of you all. Important thing is to understand what the color spaces do to the images and I think on that point there are still a lot of folks who don't really know because they don't care, as long as their prints look good.

Dan has a very specific point of view with color correction and processing in Photoshop, and given his stature, experience, and level of contributions, I respect his opinions on the matter. There's no doubt that he has developed and shared a great many tools for image processing, many of them working in the L*a*b* space.

That said, there are many aspects of image processing with which I disagree with Dan. First, he feels that selective edits (ones in which masks are used) are essentially a waste of time. All (or nearly all) of his tips involve global edits only, affecting the entire image. Second, he feels that 16-bit editing confers little to no benefits over 8-bit editing. Third, he feels that shooting RAW provides little to no benefit over shooting JPEGs. Finally, he doesn't appear to be concerned with making the most of the gamut of current inkjet printers. I won't go into the details of why I disagree with Dan on each of these points, but it's relatively easy to come up with examples for each of these points where I believe he's mistaken.

Making an 8-bit LAB edit is far more damaging to the smoothness of the image than making a 16-bit ProPhoto RGB edit ...

howlandr wrote:Like Eric, I understand the theoretical reason for avoiding a wide gamut color space for use with a narrow gamut image. Part of the theory is not just that the larger gamut means wider spacing between discrete color tones, but also that some of the ProPhoto RGB colors are purely artificial and do not represent real colors that could be seen, captured or printed. (This is called coding efficiency -- how much of the color space encodes usable colors.) And I trust that a guy like Joseph Holmes has created a highly tuned series of color spaces of varying size gamuts for a good reason.

Having said that, I've been using ProPhoto RGB exclusively for some time now and I do some fairly heavy post-processing. So far have not once encountered a tonal banding problem in a 16-bit workflow. What I have seen frequently is gamut clipping with Adobe RGB, effecting colors captured by my cameras, and printable by my printers. ProPhoto RGB lets me preserve those colors at the cost of being aware of how to work effectively with the color space.

If excessive gamut size was a problem for narrow gamut images I'd think it would create trouble given that applications like ACR, Lightroom and Lightzone have all adopted (variants of) ProPhoto RGB as their internal working color spaces. They do not give the user the option to select anything different the way Photoshop does. Of course these applications are all 16-bit internally as well.

My conclusion has been that gamut size issues are only likely to show up in an 8-bit workflow. Since I never go to 8-bit until the very end of my post-processing stages for output to print or web, my belief so far is that ProPhoto RGB does not have a downside in a workflow designed for image fidelity. The only real downside I've hit so far is not being able to fully see the gamut on current displays, but there are ways to work with that...

Certainly using 16 bit would minimize the spacing issue between specific tonal points. But while it may be hard to distinguish the change from any specific tone to its neighbor, the effect can be cumulative. For issustrative puroioses only, if we take an arbitrary measurement, and say that the difference between a specific tone and itsd nearest neighbor is 10 for Adobe RGB 1998 and 12 for ProhotoRGB. the difference of 2 may be hard to distinguish in isoaltion. But if we look at a real example, say of a sky that gradually changes in tone from light blue to medium blue, the number series would look like this for the smoothest transition of continuous tones:

Adobe RGB 1998: 10 20 30 40 50 60 70 80 90 100
Prophoto RGB: 12 24 36 48 60 72 84 96 108 120

So the cumulative effect of of wider spacing for Prophoto RGB causes the 10th point to now be 20 values different than Adobe RGB 1998.

In reality the Prophoto RGB series would look something more like this:
12 24 24 36 48 60 72 84 84 96

But now you create some banding issues.

Whether or not any of this is visible will depend on a specific image and a specific viewers ability to detect tonal changes. But I think this makes it clear that there are certainly benefits to matching a color pace to the image, as opposed to using the same color space for all images.

Ed -- I agree with Eric on this one. Dan Margulis has done a ton of hard work to contribute many great things to the imaging world... even if just by being a strongly opinionated contrarian to force others to be really sharp. But if we wanted to have a battle of dueling experts we could get into all of the equally experienced & credentialed people like Andrew Rodney and Bruce Fraser who put out very different approaches on many important matters. It doesn't make DM wrong about everything, but the fact that some of his ideas are good doesn't make him right about everything either. I use experts to identify areas of workflow that I may want to pay attention to. But at the end of the day three things make more impact to me than just the experts' face-value statements:

• - Trying to sort out the underlying theory of why things work the way they do, rather than just following an expert's blanket statement such as "always do X" or "never do Y" just because I the expert claim it's mindless to do otherwise
  - Pragmatic experience of working with a technique for my purposes, and weighing the benefits vs. the problems
  - How well the technique meshes with everything else I'm doing in my workflow

Eric has mentioned some of the short-comings of parts of DM's base level technique. In your selected quote, he claims that wide(r) gamut images don't exist, or grudgingly admits that even if one can be found it's a rare case. Well as I showed in this thread, I found one myself in the first test case I looked at and suspect my library contains many more. So they can't be all that rare. Further his use of polarizing terms like "mindless extremes" and "ham-handed", seems to me to have moved beyond simply having strong opinions based on facts, to a level of religious fervor. I don't doubt that he has good reasons for what he says, but some of his points contradict my direct experience and he seems to lace his factual info with some big dollops of rhetoric. It makes me have to work harder to mine the useful parts of his approach. This is annoying... but I digress.

My goal is to understand things, test them, and use them if pragmatic & beneficial. Part of what I wanted to accomplish in commenting in this thread has not been to be argumentative, "right", or convince people to use a technique that they don't understand or need. Rather it has been to debunk some of the simplistic anti-ProPhoto RGB sentiment that has been publicized out there and demonstrate why using it could be beneficial for some people, based on real cases and a real way to test images of one's own...

Greg -- I'm sure you know, tonal separation in itself is not a problem, as long as within the context of a given color space the colors look and reproduce equivalently. The human eye does pick out relative differences more easily than absolute differences... color shifts for any reason would be undesirable. As I've said all along we have to be aware of the trade-offs in working with something like ProPhoto RGB and not be (to borrow DM's term) "ham-handed" in applying it. The real problem would be applying a set of image adjustments that distort the color in undesirable ways. This can be just as easily done in sRGB, Adobe RGB or any other by a person who is heavy-handed or imprecise in their post-processing work.

So it comes back to knowing how to use the technique appropriately according to its strengths & weaknesses. If the choice is clip saturated colors or risk shifted colors, I'll take the latter and learn how to make more precise corrections. For example one simple thing I adopted awhile ago is making most tonal adjustments on separate layers with the blending mode set to luminosity to hold color shifts at bay. It works equally well in any color space but is potentially more important with ProPhoto RGB images because of the wide range of tonal values and saturated (even artificial) colors... in essence these factors give people more rope with which to hang themselves, so care must be taken. But really that level of care should be taken in a smaller color space anyway... distorted color is distorted color whether the gamut is wide or narrow, or the tone values are dense or spaced out.

As long as the colors are not distorted, any image in Adobe RGB or ProPhoto RGB is going to be converted back to another (usually smaller) color space for viewing onscreen or printing, anyway. So the wider tonal spacing of ProPhoto RGB is not the final story and we also have to pay attention to how it compresses back down for output.

Your suggestion of matching the color space to the image is no bad thing. If I had a complete level of discipline right now I'd be using Joseph Holmes' series of 5 DCAM color spaces, which I've mentioned before. But there are some pragmatic reasons I don't (aside from being lazy ), one of which is that I can't select any of them as the output of ACR for RAW conversion. The other of which is I haven't put into my workflow the critical step of evaluating an image's real gamut to know how large or small a color space to use for it. I do know that I have many images where color clipping may have occurred in the past, so right now I'm choosing to err on the side of caution using ProPhoto RGB as the default, and don't seem to be suffering for having done so.

Be that as it may, we still have the fact that imaging applications like ACR, Lightroom, Lightzone, etc. are wired internally for ProPhoto RGB no matter what color space you convert an image file into. HDR applications, which I'm working with, introduce a further challenge by going to a 32-bit tonal spectrum as the "working space" (although HDR image formats are not really color managed yet in the same way). So we are going to have to come to grips with working in big spaces with high resolution for tones, and gamuts involving saturated colors. IMO it's eminently doable, for those who want that extra bit of kick.

As a side note I'd be interested to see a real test case of the scenario you outlined here with a blue sky gradient for example. I don't know if you have time to create one. I may have a go at it if time permits in the next couple of days...

WOW Royce! Thanks a lot for your effort to explain this!