How to manage images for web and print efficiently?

Hello guys.

I have my monitor calibrated to D65 because the target devices for my photos are computers and smartphones. Pretty common I suppose.

I alter my images so that they look realistic using D65 display profile.

If I decide to print several of them, I imagine that printed photos will be warmer in comparison to their look on D65 monitor. Right?

Is there such a feature or a workflow which can convert image looking good in D65 to image having precisely the same look as in printed version (D50)?

I suppose this have something to do with working profile. Intuitively something like “transform raw image values with values coming from the difference between D65 and D50”.

Thanks!

EDIT: I just read that specifying wide gamut output profile like RT_Large_gsRGB will help in the case the printer supports it as well. https://rawpedia.rawtherapee.com/Color_Management#Output_Profile

To make it short: the different profiles are used to map colors with reference to a set of primary colors specified in the color space present in the profile. And they also have instructions on how to convert from one profile to another. Basically you have to make sure that you choose a profile (color space) big enough to include all the colors present in your images, both on display or in print.

The setting that takes care about converting or not between different white points is the Rendering Intent. Many people say that for photography you should use the Perceptual rendering intent. I use the Relative Colorimetric rendering intent because is more faithful with colors, but at the expense of loosing the most saturated (and usually clipped) colors.

So keeping ProPhoto as a working profile and RT_Large_gsRGB as an output profile will enable me to keep as much colors as possible for both printer and monitor. Subsequently, experimenting with Rendering intents will show me which intent is best for printing. Right?

Differently said, if I change images while using D65 (display), ProPhoto (working) and then output them using RT_Large_gsRGB profile & proper Rendering Intent and finally make wide gamut print of the exported photo, will I see (more or less) the same colors in print when compared to colors in D65 color space?

EDIT: provided I use wide gamut monitor as well :slight_smile:

Generally speaking, yes. That’s not the only possible combination, but it may suit you.

You may have to often soft-proof your image while editing it, though

Hm, ok. Thank you very much for quick responses.

While profiles in most cases take care of the colors in a satisfactory way, the harder part is managing the differences in contrast between raw captured scene contrast (~1:10000 for a contrasty landscape and decent camera), display contrast (~1:1000 for IPS displays) and print contrast (~1:100). Soft-proofing can help to visualize what contrast can be expected.

This mapping of captured contrast into the final image is part of the creative process, much less ‘mechanical’ than the handling of color, for most scenes. In rawtherapee, I usually first create a screen version, export it together with the ‘screen’ profile and then continue on for a print version, which I again export together with the profile. This way I can rework any of the two if I find a need later - and I know which profile to load and continue from.

2 Likes

Please read this if you start working with printers and soft-proofing: http://rawpedia.rawtherapee.com/Preferences#Color_Management_Tab

You really want to use soft-proofing, because generally you have no idea about the shape and size of the destination color space. You need to have an ICC profile that matches your printer to predict how your image is going to look like. You can either obtain a colorimeter and make such a profile yourself, or contact your local printing facility and ask them for a profile for their printer.

Note that simply setting your output profile to some large color space, doesn’t mean your image will look good on all media. If your desktop application is not properly color managed, it may assume an sRGB profile. This really screws up your colors. The same goes for printing.
To play things safe, choose an output profile for your image that matches your inteded medium.

2 Likes

This whole topic starts to seem exponentially complex to me :smiley:

Firstly, thank you guys for your time. Really.

@dabbler
Hm, if I understood your workflow correctly, firstly you make changes using D65 display profile while using soft-proofing set to output profile you chose for “screen” type (e.g. one of the sRGB versions) and export to/with that profile. Secondly, you make changes using D50 display profile while using soft-proofing set to chosen printer’s profile + export. And if those two exported versions need further corrections, you just manipulate them separately using D65 or D50 monitor profiles accordingly. Right?

@Thanatomanic @dabbler
So to summarize:

  • I really need soft-proofing so I will see previews how things look like in different color spaces of my final mediums.
  • By looking at the previews I can change photo specifically to those mediums separately.
  • I really should use sRGB or RT’s version of sRGB as output profile for “screen” type to cover non-color-managed screens, which is the majority of smartphones, monitors, laptops.
  • I should get my printer’s color profile.

Questions:

  • Is there any use case in which soft-proofing is not used? Because essentially, I always have to manipulate the image colors and contrast according to the final medium.
  • What if I do not know the final medium, is it worth to make any preliminary changes without this information?
  • I have calibrated my monitors to both D65 and D50. Not sure if Rawtherapee has access to this information in runtime. Especially, when I move windows here and there. What does it mean to use soft-proofing set to printer’s profile while looking at the D65 monitor? Or should I switch to D50 profile for this particular task?
  • May I ask @dabbler which format you chose to store those two exported files? Simply JPEG?

PS: Just a fun fact. I have done a color calibration with Pantone Huey of my external monitor and found out that it does not cover that much of sRGB as my laptop. My expectation was quite the opposite.

Monitor: Dell P2417H
Laptop: Latitude E5570

I don’t use different color profiles for web and print. Human vision is very adaptable to different whitepoints. I use a 27" monitor, which allows me to have a fair amount of neutral frame (e.g. grey tool palettes) around my pictures when editing, so it doesn’t matter too much, which specific whitepoint I use. Read-on below for my explanation why, with the disclaimer that I’m just a physicist by education but by no means a color management expert.

BUT, my point earlier was that once you have some reasonable color management in place, irrespective of white point, you actually should pay attention to image contrast next, because it will have a vastly bigger impact on your prints than subtleties, which potentially are addressed by profiles with different whitepoints. Managing contrast includes handling gamut limits in bright areas. A convincing sunset ‘sky on fire’ is a lot more difficult to create in print than on screen, due to the limited contrast and the saturation limits that go along with it.

Back to color profiles, the rationale for 6500K for web editing is likely that that’s close to most screens native (=brightest) white point, similar to overcast daylight. Therefore, you’ll have a similar viewing impression as many others, when you ignore the influence of the current viewing environment. If you use a 6500K profile in a room lit with popular (here in Germany) 2700K ‘warm’ household lamps, your screen will look blueish, not white, unless it fills a really large portion of your field-of-view. So probably your edits will end up with a rather warm touch. On a small screen, a dark room will, while maybe color-balance wise OK, on the other hand lead to rather dark edits, unless there is a lot of tonal ‘reference’ material placed on the screen around it. Editing in a room with daylight would likely result in more balanced edits as long as it isn’t too bright.

Editing for printing is basically no different than for web use, when it comes to color balance. I have successfully edited a lot of pictures for printing with my monitor’s native white point (close to 6500K). Color profiles take care of mapping colors in the reference color space (CIE, modeling the average human color perception) to a specified viewing condition, including different white points. This allows to transform color values between them, which will then emit the best matching (according to the rendering intent) stimulus, achieving the closest color perception despite the different viewing conditions.

The major difference arises when you actually look at prints, which are lit by some defined light spectrum, to check them or a color chart or picture for reference in the same environment as you use your monitor for pictures on-screen. Then you do need to match those reflective and emissive viewing conditions simultaneously as closely as possible, so that they generate consistent stimulus. This is where e.g. D50 lamps, color neutral booths, and screen profiles with matching white points and brightness come into play. If you print yourself, it may be worthwhile to create a screen profile closely matching your print viewing conditions, otherwise taking your ‘average’ working environment approximately into account should be sufficient.

No. That’s the reason we have chromatic adaptation.

Not sure what you mean by that. A color profile is just a description of a color space. A color space is a 3D vector space, where primaries define the base vectors. When you digitally change the temperature, you are changing the “color” of the white point, and then the chromatic adaptation will change the RGB primaries accordingly, to compensate.

A D50 print will not be warmer on average than a D65 because your perceptual system adapts quickely to illuminant. However, if you compare side to side the both, indeed one will look warm and the other will look cold. This why pre-print monitors are calibrated for D50. But the relevance of comparing a backlit emissive medium with a front-lit reflective medium is still to be proved. For That reason, “soft-proofing” fails to do its simulation job most of the time, so take it as a toy to keep the nerds entertained.

But aside from the direct side-to-side comparison, everything should be fine provided your color management system works properly and has the right profile for the right medium.

FLOSS guys should really stop freaking out with color profiles and such. These problems have been solved 30 years ago.

I wouldn’t entirely dismiss it. Short of making test prints using the target printer, which may not be practical in terms of turn-around time and cost when you order prints, soft proofing is the next best thing to visualize where you need to further work on your images. Out of gamut indicators often fail to serve that purpose for me, because I e.g. don’t care too much if colors are slightly out of gamut in very dark areas as long as tonal differences are subjectively OK. For judging where I further need to reduce global contrast and work on local contrast, I find it immensely useful - much more visual than histograms and numbers.

After a few iterations, I have learned that I get just a little more depth in dark areas than the soft-proofing profiles of my printing shop simulates and that it e.g. matches the print appearance for matte prints at 30-40 degrees viewing angle under our typical semi-diffuse room lights. Knowing that, I can now reasonably pre-visualize how the prints will look like, which I’m totally unable to based on the ‘bare’ monitor picture alone. It’s enough to most times order larger formats without a round of proof prints first.

1 Like

That means the black point adaptation performed by the CMS used to convert to print RGB space is failing. It could be either because the CMS doesn’t apply it properly, or the ICC profile is wrong.

The whole point of the ICC toolchain is specifically to avoid these half-artistic color corrective dabblings, and bring reliability in the process.

After reading up on it, I don’t think so. In fact I think it should be fairly correct, because the standard measurement geometry according to ISO 13655:2009 for prints is 45º/0º or 0º/45º (light and sensor angle), see e.g. https://www5.konicaminolta.eu/tr/oelcuem-cihazlari/online-kuetuephane/renk-oelcuemue/colour/iso13655-demystified.html - so my observation that it matches well at an angle, which isn’t necessarily the viewing angle with the highest contrast, is in line with that.

I don’t understand what you are comparing here. The 0°/45° is only a mean to remove specular reflections on the paper substrate, that could produce hotspots. You have probably experienced that in museums, where glossy oil paintings are exposed in front of windows… sometimes, all you see is the reflection of the window over the canvas, it’s quite annoying. But in the other way, when the window is on the side of the painting (90°), all you see is the bumps of the canvas texture.

So that 0°/45° thing is the optimum that avoids parasite harsh reflections of the light source on the (more or less) glossy medium and ensure evenness of the light field, without emphasizing the substrate texture. This shouldn’t affect your perception and is surely not related to the black density.

Every attempt at matching “a viewing result from a certain angle” on the sole density of the emitted black is a nasty hack because, again, you cannot simulate a reflective print (even more if you add non-lambertian lighting in the mix) on an emissive LED panel.

The very reason of the annular 45° lighting in your reference is to cast away the non-lambertian effects, and create viewing conditions that approximate the emissive display. So it’s the other way around.

You may have gotten that one backwards - viewing the final print from 45 degrees when lit from the top is simulated by the softproof remarkably well (the angle on the monitor doesn’t affect perception much) - for the purpose of using it as a pre-visualizing aid for the print. I do not claim that the angular dependency is in any way simulated by the softproof, just the appearance of the print at a certain, and relevant, viewing angle and lighting direction.

And there is a rather simple explanation for that angle and the assumed blackpoint problem: While in theory

is the motivation for that ISO definition, it’s not a silver bullet for all media. Notice that I wrote that this is a matte surface print. It does have a fairly wide angle soft gloss, which has not completely vanished at 45 degrees, even though it’s far from pronounced. This can visually be explored with the help of a torch.

For illustration (not to be mistaken for a quantitative assessment), I have taken two pictures of such a print: Camera perpendicular to the print and lit by a flash angled at 45° (measured), pointing directly to the center of the crop below. Then I lowered the flash to ~30 degrees (estimated) and took a series of exposures, increasing flash power in 1/3EV steps. Equal brightness in the mid-tones and highlights was between +1/3 and +2/3, so I used +1/3 with an exposure correction of 0.12EV in darktable to bring them even closer together. No post-processing as such otherwise, just identical whitepoint, crop, input/output profile, demosaic enabled otherwise and resized to half size.

First 45°


then 30°

If you compare those, the differences are subtle, but there is a little better contrast abd detail in the dark rock next to the house with the shallower light.

So the explanation is simply in the fairly finely but deeply textured paper surface, still causing some glow at 45°.

Notice that I did swap to a 45°x0° arrangement from 0°x45° for practical reasons, but all the above discussion equally applies.

So, in practice, with all the fundamental imperfections and limitations, I find softproofing to be an immensely helpful and reasonably accurate tool.

But a torch has never been in the mix. You need to use a diffuse light, large and even, from all directions around the print. Of course with a punctual light source, you can show whatever you want.

Same remark. A flash is not suitable. Of course it will enhance paper texture. And that is not something you should bother with while adjusting global contrast.

Anyway, do what you will, but my official statement is soft-proofing doesn’t work reliably in general and the current state of research suggests that trying to emulate reflective materials on an emissive panel is doomed to fail. (Yes, you can find that one example where it works ok, but a stopped clock is accurate twice a day).

Even 3D softproofing emulating paper surface works randomly : https://scholarworks.rit.edu/cgi/viewcontent.cgi?referer=https://scholar.google.fr/&httpsredir=1&article=1159&context=other

Once again, it seems that the solution might lie in spectral domain : https://www.researchgate.net/profile/Theo_Phan_Van_Song/publication/311592751_Towards_Spectral_Prediction_of_25D_Prints_for_Soft-Proofing_Applications/links/5a9559d50f7e9ba429718e51/Towards-Spectral-Prediction-of-25D-Prints-for-Soft-Proofing-Applications.pdf (but the results have not been submitted to a test panel).