Clipping out-of-gamut colours

[RobertArdill]

Hi,

I usually prefer to use Relative Colorimetric for printing, but the clipping of out-of-gamut colours can be a real nuisance; changing to Perceptual may not be an option because the colours may be shifted too much.

I don’t much like trying to tweak the saturation / hue … of the out-of-gamut pixels because I find it difficult to do. I usually end up with a result that’s often not at all good.

It occurred to me that what might work is a combination of Relative and Perceptual, with the out-of-gamut colours in Perceptual (which would bring them back into gamut) with the rest in Relative.

So, I tried the following on an image: I took a copy and converted it to the print profile using Perceptual. I then converted it back to Relative (with BPC off). Then I copied it over the original and painted over a black mask to get rid of the out-of-gamut colours (in other words, painted over the Gamut Warning red). I did this quite roughly, not being too careful.

To finish off, I turned off the gamut warning and toggled the ‘Perceptual’ layer on and off to see which bits I preferred in Perceptual and which in Relative (with soft-proofing on, of course). I then masked out the areas that I wanted to be Relative.

It’s also possible to vary the opacity of the ‘Perceptual’ layer, to get a blend of the two – this is softer but still removes or softens the Relative clipping.

I haven’t come across this technique before, but it does seem to me that it could be useful in some situations. It seems complicated, but it's easy enough to write an action to create the 'Perceptual' layer, so then it's only a question of painting on a mask (and tweaking the layer opacity, possibly).

Here’s an example with the Cambridge in Colour sample image:



I would be interested in what you think of this technique.

Robert

[Black Pearl]

What do you generally take and print pictures of?

[chauncey]

I'm assuming that you do your own printing...I might suggest that you overly engrossed in the process rather than the final product. A video that I once watched instructed "when in doubt, make a print" and, judge your technique on the results of the print.

[Manfred M]

Hi Robert – this thought of yours is quite in line with your previous posts. You seem to be so wrapped up in minutiae that you are missing the big picture.

Rendering intents have a single purpose, which is to provide different models of taking an image (or for that matter graphic art) and bringing the colours / shades in it down to something your printer can reproduce. As we are primarily looking at photography here, let’s stick to this part of the field of graphic arts.

A higher end modern camera is capable of capturing around 12 stops of exposure. Your computer screen can only handle around 5 to 6 stops and your printer / paper / ink combination can reproduce down around 4 stops, so the rendering intents are used to compress the exposure range down to something that your printer can output.

The second part of this issue is related to the number of distinct colours your printer can reproduce. Many people sneer at the lowly sRGB gamut that can only reproduce 16 million distinct shades, but this outshines your high end photo printer by many orders of magnitude. The best estimates I’ve seen suggest even high end printers are likely reproducing somewhat a bit more than 500,000 distinct shades, so again, your rendering intent has to take the colour input and map it to something that your printer can reproduce.

I don’t get too concerned about blacks when printing; my Epson 3880 does have a nice black cartridge, so any loss of shadow detail is taken care of with nice heavy, large black dots put down at 360 dpi. Whites, on the other hand, I do worry about as this is produced by not using any ink at all, so there will be unattractive white spots at any value that has a white value of greater than (250, 250, 250), so I do manage things to ensure I do spray some ink even in areas where I have lost highlight detail in the working image (by setting my maximum white point in the 245 to 249 range.

I’m not sure why you are getting into this level of detail; it makes ZERO difference in the final printed product. The human eye / brain visual system lets us ignore these fine differences in detail you are so wrapped up in.

[RobertArdill]

What do you generally take and print pictures of?

Landscape - varies from low light to brilliant sunsets, so very different hues and saturation.

[RobertArdill]

I'm assuming that you do your own printing...I might suggest that you overly engrossed in the process rather than the final product. A video that I once watched instructed "when in doubt, make a print" and, judge your technique on the results of the print.

Hi Chauncey ... I wish Ansel Adams was still around so you could say that to him!! I've spent years printing as you suggest I should ... with mixed results. Good sometimes, not so good at others. Now that I am paying attention to the minutae (gamut, tonal range, monitor, printer, paper ...) my results are not only much better, but much more consistent. I almost always get exactly the print that I aimed for, which is very much the final product for me since this is (mostly) what my customers buy and pay me for. So I have to say that I completely disagree with you.

As a significant additional benefit, when you do go into the minutae, you have to understand what is under the hood, what makes a good profile and what doesn't, why this paper is much better than that paper for this particular print, the effect that a good monitor with good calibration has, why it's important to view your print under controlled conditions, and then view under different lighting, the effect of optical brighteners etc., etc. So it does take some time, but the time is well spent IMO. Once you've spent the time and understand how to make the best use of your equipment, given its limitations, well then it all becomes quite easy and it doesn't take the very time-consuming and expensive trial-and-error, print and reprint method that you suggest as being the best approach.

But I do agree that at the end of the day the print is the final arbiter .

[RobertArdill]

GrumpyDiver;421459]Hi Robert – this thought of yours is quite in line with your previous posts. You seem to be so wrapped up in minutiae that you are missing the big picture.

Rendering intents have a single purpose, which is to provide different models of taking an image (or for that matter graphic art) and bringing the colours / shades in it down to something your printer can reproduce. As we are primarily looking at photography here, let’s stick to this part of the field of graphic arts.

A higher end modern camera is capable of capturing around 12 stops of exposure. Your computer screen can only handle around 5 to 6 stops and your printer / paper / ink combination can reproduce down around 4 stops, so the rendering intents are used to compress the exposure range down to something that your printer can output.

The second part of this issue is related to the number of distinct colours your printer can reproduce. Many people sneer at the lowly sRGB gamut that can only reproduce 16 million distinct shades, but this outshines your high end photo printer by many orders of magnitude. The best estimates I’ve seen suggest even high end printers are likely reproducing somewhat a bit more than 500,000 distinct shades, so again, your rendering intent has to take the colour input and map it to something that your printer can reproduce.

I don’t get too concerned about blacks when printing; my Epson 3880 does have a nice black cartridge, so any loss of shadow detail is taken care of with nice heavy, large black dots put down at 360 dpi. Whites, on the other hand, I do worry about as this is produced by not using any ink at all, so there will be unattractive white spots at any value that has a white value of greater than (250, 250, 250), so I do manage things to ensure I do spray some ink even in areas where I have lost highlight detail in the working image (by setting my maximum white point in the 245 to 249 range.

I’m not sure why you are getting into this level of detail; it makes ZERO difference in the final printed product. The human eye / brain visual system lets us ignore these fine differences in detail you are so wrapped up in.

Hi Manfred,

I’m not sure what the big picture is, but my feeling is that in painting and photography (I am half-and-half a painter and photographer) that the big picture is what you put up on the wall when it’s all over. Paints have a far smaller gamut (generally speaking at least) than printer inks, so in painting it’s even more important to understand how hues and tones interact, because it’s often in the relative that you get the effect. If you don’t pay attention to detail then what you may end up with is an image that is flat and dull. Small changes can make big differences.

So, to your point about camera and printer/paper stops etc. Just so. Even though you are getting into the minutiae here , it is important to understand this, because it is by understanding why the print just doesn’t look as good as the scene you photographed that you can make adjustments to it that can make it look better than the scene you photographed. This is where skill and art and understanding come into play, and it is one aspect of photography that helps to make the difference between a holiday snap and a stunning print.

Regarding the printer gamut versus sRGB and the different numbers of hues and shades that each can reproduce: the same argument applies here. However, in addition, the printer gamut in the case of a paper like Canson Platine or Photo HighGloss on a printer like the HPZ3100 is WAY bigger than sRGB. This doesn’t mean that the printer can produce a greater number of individual shades and hues than sRGB on a good monitor – it means that the printer can print colours that cannot be seen on sRGB on a good monitor. As long as there isn’t posterisation, the gradation between colours isn’t all that important IMO, because the eye isn’t able to distinguish between very fine differences in colour (typically, very few people will be able to see the difference between two colours with a dE of 1, which is only 1 in 200, not 1 in millions).

As for blacks. No, I don’t agree at all. Shadow detail is one of the things that make a big difference to an image, whether viewed on a monitor or printed. If you know that your printer cannot print darker than say L:5 or RGB:24:24:24 (which is very good as this corresponds to a DMax of 2.26) then you should not be sending data below this level to it because if you do you will get a solid black (unless you want solid black, of course ). If the best DMax you can get is 1.8, say, then anything below RGB:40:40:40 will be pure black (or perhaps I should say printer black), so now there is real potential for problems! Solid surfaces with no tone or hue change at all really do need to be avoided – they look terrible. Even a slight tonal variation makes a big difference.

I mostly agree with you about the whites – again, clipping is generally not a good thing, especially if it is over a fairly big surface (like part of a cloud, say), because it makes the surface flat and lacking in quality. However, sometimes it is good to print pure white (a specular highlight, for example), in the same way that sometimes it’s good to print pure black.

So why, as you ask, am I getting into all of this level of detail? Firstly because I find it interesting; secondly, and more importantly, because I think it makes a big difference to the image. The human eye does indeed have amazing abilities – but it doesn’t have the ability to render a sow’s ear into a silk purse. It won’t add contrast to a flat image, put colour where there is none, correct a hue. What we, as photographers, need, is to understand what the eye can and cannot do and what we can and cannot do with our photographs. For example, we can add local contrast to an image (the eye can’t do that) – and this can make a very big difference to what the eye perceives.

I do agree that there are some things that are just not worth doing. For example, some photographers profile their cameras for every lighting condition they are likely to encounter (or bring a ColorChecker with them so they can later produce the profile for the particular condition they have encountered for this particular shot). I’ve checked my camera (a Canon 1DSIII) and I absolutely do not think it’s worth the trouble for my type of photography, so I don’t bother. But at least, having checked, I now know that it’s not worth the bother: the improvement that I might possibly get is so marginal as to be irrelevant, to my eyes at least.

Now I would like to ask you a question. In response to the two posts I’ve made on what I hoped to be a forum open to discussing ideas and techniques I’ve encountered 100% opposition. Not one of you who have replied has attempted to see anything useful or positive in what I’ve posted, and it seems to me that you also have not tried out my suggestions (but perhaps I’m wrong). Is this because I’m the new kid on the street and you’re putting me in my place? (in which case I won’t be on this street for much longer). Or do you genuinely think I’m being totally obsessive-compulsive and need a reality check?

The reasons for my posts were two-fold: to get constructive feedback and suggestions; and to offer to others what I think may be useful techniques. I have to say I’m a bit disappointed by the response … but hey, that’s life .

[Manfred M]

I quite agree with you in terms of the range of what a printer can produce (I suspect it is likely a bit more than AdobeRGB), so we are looking at limited colours within a 3-D envelop that is significantly greater than sRGB.

As for why the printer cannot 100% accurately reproduce the colours in a painting; that is hardly surprising. The only type of painting that more or less works the same way as an ink jet printer is water colours. The pigment lies flat on the paper substrate, with some absorbed on the fibre base while other parts on the surface. Any 3-D effects are largely due to the paper's texture.

The moment you go to an acrylic or traditional oil, you have a thin, but very real 3-D object with significant ridges and textures, especially if you use a palette knife. Even traditional brushes leave siginicant texture. Add to that the filler material used to reflect light in the paint (clays and other fillers), so you get light reflecting off in different directions. I would not be at all surprised of metamerism does not come into play an cause issues; different pigments will have different spectral responses.

[Daveart]

Hi I don't know what printers you guys use but I have a a3+ canon printer and the gamut range is far more than my screen, the canon print uses 16 bit printing which is greater than a monitor which only displays 8-10 bit at a time and I can send pure CMYK profiled direct to the printer, plus it prints at 9600 dpi as apposed to 4800 dpi of Epson printers I used to use Epson. I am an artist the width of gamut in painting is only limited by the artists knowledge, you can probably produce something like a trillion colours, which is probably around 64 bit colour depth.

what the computers come sup with in the gamut warning is a colour depth gamut of 8-10 bit, so anything beyound this is unprodictable colour correction due to srgb colour range is your print can print a CMYK profiled image send it to your printer at that unticking Photoshop management and let the printer manger the colours for you you will get a more accurante colour printed image.

Dave

[DanK]

Now I would like to ask you a question. In response to the two posts I’ve made on what I hoped to be a forum open to discussing ideas and techniques I’ve encountered 100% opposition. Not one of you who have replied has attempted to see anything useful or positive in what I’ve posted, and it seems to me that you also have not tried out my suggestions (but perhaps I’m wrong). Is this because I’m the new kid on the street and you’re putting me in my place? (in which case I won’t be on this street for much longer). Or do you genuinely think I’m being totally obsessive-compulsive and need a reality check?

I haven't responded only because I know too little. My printing skills are far more rudimentary, and I have to say that I really struggle with some out of gamut colors with the printer and papers I use.

However, to give others their due, it is often the case that folks post concerns that in the end don't matter. For example, it is easy to find people posting on the web who insist that photos will be ruined by diffraction if you go two stops smaller than the optimal. In this case, however, you've made a strong case that these are details that do matter.

I hope you continue to post, because I will learn from it.

[Daveart]

Hi Dan, I understand your frustration if you can get your head round the fact that your screen uses RGB red green blue colours and your printer has CMYK which is cyan, magenta, Yellow and black the computer has to covert the rgb to CYMK as you can see you have one more depth in the printer that your screen has each has 256 tones of each colour so a screen has 256x256x256 that is how many colour it can show at and one time where as a print has 256x256x256x256 which is far greater.

If you calibrate your screen and printer you will have them both match the colour depth so as your printer will print to how it looks on the screen.

Dave

[RobertArdill]

[

I haven't responded only because I know too little. My printing skills are far more rudimentary, and I have to say that I really struggle with some out of gamut colors with the printer and papers I use.

However, to give others their due, it is often the case that folks post concerns that in the end don't matter. For example, it is easy to find people posting on the web who insist that photos will be ruined by diffraction if you go two stops smaller than the optimal. In this case, however, you've made a strong case that these are details that do matter.

I hope you continue to post, because I will learn from it.

Hi Dan,

Thanks for your reply. Believe me when I say that I too have struggled for a long time (and no doubt will continue to) with all of the technology involved in getting an image from camera to print without being disappointed in the end (or surprised at how great it looks, which is almost as baffling!).

The whole idea of colour management is to cater for the (very big) differences between the camera, monitor, printer, scanner, projector, printing press, lighting etc. (not to mention the eye ... in all its complexity and variability). That is manages to do so at all is pretty amazing ... but it's not surprising that if we get things a bit wrong along the way that it then really doesn't work, and we end up being disappointed and frustrated.

The biggest mistake I’ve made (and no doubt so have others), is to place too much emphasis on the newest gizmo and not enough on getting to know what I do have really well. So I’ve kept trying different papers, for example, without ever really getting to know one really well.

I guess this is something that just takes time, a lot of time, and perhaps the important thing is not to lose one’s passion along the way. If we keep moving ahead then there is always the gratification of achieving something that we couldn’t have done last year.

Bits and bytes and gamuts and profiles and all that jazz – well, it’s just the nuts and bolts. Still, if we want to make the best of what we have at our disposal, we do need to understand them, because if we don’t we’re like a rudderless boat.

So – is it important to have some idea of what colours might be clipped when you print using a Relative intent? or be able to see what happens when you use a Perceptual intent that may shift all the hues (in a way that you have no control of)?

Well, you can answer that question as well as I can, of course. Then: what about the other question: so, you can see what’s happening and you don’t like it, what can you do about it? Is that an important question?

Robert

[RobertArdill]

I am an artist the width of gamut in painting is only limited by the artists knowledge, you can probably produce something like a trillion colours, which is probably around 64 bit colour depth.

Dave

Hi Dave,

I really seem to have managed not to get my point across at all. My post really had nothing to do with how wide or narrow the gamut of a printer or monitor or camera is – it just had to do with what to do if the output device messes up the image because it’s screwing around with the colours because it can’t handle them. If you want a nice gradation of hue or tone and you get a sharp transition instead … then you might want to see if you can do something to correct that problem.

As for painting versus photography: when I paint I’m more interested in limiting the gamut I use than in extending it to 64-bit depth! In fact, I think this also generally applies to photography: more is often less (or vice versa) … which is perhaps why so many great photographers had little interest in colour. So perhaps a more useful way to phrase my original post is: How can you effectively limit the gamut of your image to achieve the effect you want?

You say: "What the computers come sup with in the gamut warning is a colour depth gamut of 8-10 bit, so anything beyound this is unprodictable colour correction due to srgb colour range is your print can print a CMYK profiled image send it to your printer at that unticking Photoshop management and let the printer manger the colours for you you will get a more accurante colour printed image."

You are right when you say that if you do not have a colour managed set-up and you are working in sRGB that you are better off letting the printer look after things, because you’re more likely to mess things up otherwise.

But … that would be like a tinned fruit-salad compared to freshly-sliced fruit. OK, if you like tinned, I guess .

[Black Pearl]

I'm not sure its a case of you not getting your point across I think its more a case of members feeling you may have become fixated on one tiny minutiae within the entire photographic process and could well have lost sight of the bigger picture.

That of enjoying taking pictures and (hopefully) making some money along the way.

I understand what you are trying to achieve but would it make any difference to your clients? Have any of them at any point returned a print or possibly refused to buy a print because they felt the colours were wrong?

I know you feel there are issues but do they matter?
Could you not spend your time in better more productive and ultimately more financially rewarding ways?

[Daveart]

Hi Robert, I was trying to make a point as some one said art has less colour range than photography, it is only limited by the artist, I. Do understand what you were saying I also, wanted to show the short comings of digital photography and the only way to get around this so as not to get unwanted colours coming into the print is by calibration of both screen and printer or send a CMYK profiled image to the printer if your printer permits this then you should not get unwanted colour bleed through translation.

Dave

[DanK]

Dave,

Hi Dan, I understand your frustration if you can get your head round the fact that your screen uses RGB red green blue colours and your printer has CMYK which is cyan, magenta, Yellow and black the computer has to covert the rgb to CYMK as you can see you have one more depth in the printer that your screen has each has 256 tones of each colour so a screen has 256x256x256 that is how many colour it can show at and one time where as a print has 256x256x256x256 which is far greater.

If you calibrate your screen and printer you will have them both match the colour depth so as your printer will print to how it looks on the screen.

I have a color-managed workflow and use ICC profiles provided by the manufacturers of my papers. However, even with a relatively high-gamut paper, some colors that are within the sRGB space on my (not very expensive) monitor are out of gamut in prints, and it can be very frustrating (at least to me) to try to get a reasonable rendering. For example, this often happens with certain reds. It happened a week ago, when I printed a shot of someone running the Burlington marathon. He was wearing an intensely red shirt. It displayed fine in sRGB, but when I printed it, after soft proofing, it just wasn't right. I suspect many people wouldn't have noticed, but I did, and it bothered me.

Robert, you wrote:

So I’ve kept trying different papers, for example, without ever really getting to know one really well.

That's a valuable point. I have to admit that I have often thought I must be missing something important by not trying even more papers than I have, but it is probably a good idea to settle on a few and really learn them. I have been using two as my primary papers for a while, Moab Lasal Exhibition Luster and Red River Polar Matte, and I am gradually becoming more comfortable with them.

Dan

[Colin Southern]

Hi Robert,

My suggestion (and I don't say this very often!) is to pop across to http://www.luminous-landscape.com, and post your question in the direction of Jeff Schewe (whom I'm sure you'll be well familiar with by now) (or Andrew Rodney); I think the two of you would have very meaningful conversations on this topic.

[RobertArdill]

I'm not sure its a case of you not getting your point across I think its more a case of members feeling you may have become fixated on one tiny minutiae within the entire photographic process and could well have lost sight of the bigger picture.

That of enjoying taking pictures and (hopefully) making some money along the way.

I understand what you are trying to achieve but would it make any difference to your clients? Have any of them at any point returned a print or possibly refused to buy a print because they felt the colours were wrong?

I know you feel there are issues but do they matter?
Could you not spend your time in better more productive and ultimately more financially rewarding ways?

Hi Robin,

Actually, I have the most fun with my photography. My wife and I are both photographers and we take several trips every year around Ireland (and also France and, to a lesser extent, Italy ... and I've just come back from a 3-week trip to Mauritius). I'm one of these people who just doesn't do things he doesn't like (or has great difficulty in doing): the day I stop enjoying photography is the day I stop doing it. So your point about enjoying taking pictures is certainly very close to my heart, and if I thought that getting stuck in the minutiae was a risk for me, then I would get out of there quick.

But of course one can enjoy taking photographs and also enjoy getting the most out of them, producing prints that one is really happy with, seeing the pleasure of a customer. One of the things that is wonderful about photography (not to say that this isn't true of most things, in reality) is that there is always more to learn; and there is pleasure for me in finding a new way of getting rid of an old problem.

I also find that as I spend more time with photography that I get more discriminating, and something that seemed fine 5 years ago, no longer does. I become more careful with framing, exposure, light ... all of these things.

Have customers complained about the colours in my prints? No. But I would not be much of a photographer if my technique was to wait for a customer's complaint before trying to correct an error, IMO.

What puzzles me a bit about some of the responses to my posts is that some of you seem to feel that OK is good enough and why bother to try to improve on good enough?

I don't think spending time on minutiae is an unrewarding or unproductive thing to do. Try it ... you will be amazed at what you find out and what you learn - and your photography will improve and you will get more fun out of it.

Robert

[RobertArdill]

Hi Robert, I was trying to make a point as some one said art has less colour range than photography, it is only limited by the artist, I. Do understand what you were saying I also, wanted to show the short comings of digital photography and the only way to get around this so as not to get unwanted colours coming into the print is by calibration of both screen and printer or send a CMYK profiled image to the printer if your printer permits this then you should not get unwanted colour bleed through translation.

Dave

Hi Dave,

I tend to be a bit direct (putting it politely ) so I hope you haven't taken offense at what I said. As a painter you will know that doing something like having an area of low saturation near one with higher saturation emphasizes the higher saturation - it's a very good trick to extend the apparent gamut. Same goes for photography, of course, as it is also a visual art.

I entirely agree with you that if you want to print properly you do need a calibrated display and printer. It's simply an essential step along the way and if you leave it out then the chances of getting the print that you wanted is low. Then, using some of the tools like soft-proofing will help to see where colors can get messed up: since your monitor and printer are properly calibrated so you can now rely (to a large extent) on your soft-proof.

As for sending the image in CMYK to your printer - well, I've only ever done this with out-of-house printers as my in-house printer is RGB only. But, yes, again you are relying on the printer to be properly calibrated and profiled, so as long as your image is OK, the print should be too.

Robert

[Manfred M]

It occurred to me that what might work is a combination of Relative and Perceptual, with the out-of-gamut colours in Perceptual (which would bring them back into gamut) with the rest in Relative.

Sorry; I've been rather busy, so have not had a chance to make comments on what this means technicallly.

The math says you are creating output with colours that are totally "false", but if you like the results, who am I to argue. After all I fully advocate getting a pleasing look over 100% accuracy in most work.

In Perceptual all of the colours are shifted to bring the outliers into gamut. In Relative, the out of gamut outliers are corrected by the non-outliers are left alone. We don't know exactly how this is done because the software and printer manufacturers do not publish their algorithms; but I suspect reality is a bit more complex, especially in how Relative is implemented.

With your technique you are preserving neither rendering intent. You are doing this by taking the value of every pixel and finding the mean based on both rendering intents and outputing what amounts to an "false" colour for each and every pixel. Nicely said both in gamut and out of gamut colours will all be incorrectly rendered using this technique.

You've spent a lot of time writing about the importance of colour accuracy in your work, yet the technique you are advocating goes in the opposite direction. Am I missing something here?