Luminous Landscape Forum
Raw & Post Processing, Printing => Colour Management => Topic started by: yalag on August 19, 2014, 10:11:34 pm
-
The explanation is usually that most of internet works over sRGB. But does anyone know how export as sRGB work in applications like LR/PS? Do the colors actually get remapped to sRGB (clipped) or is it still in its original gamut but with sRGB profile tagged?
I ask because I did an experiment where I took an image from my desktop that is attached to wide gamut monitor (adobe rgb) and tried to soft proof it against sRGB. I basically tone down a yellow flower that is out of gamut then exported it as jpeg (with relative intent).
I then saved another jpeg from the original image without any adjustments to sRGB space.
I then took these 2 jpegs over to my macbook pro (with a monitor gamut closer to sRGB) and to my surprise that the non adjusted image has a lot more saturation. I was expecting them to be very close. Why is this?
I then did a third image, which is the original image exported as adobe rgb. To my surprise again, it's very close to the second image. But it's still a touch more saturated. I expected it to be the same. In this case, is it a good idea to export as sRGB?
The left is the adjusted one, the right is no adjustments (srgb), both shown on the macbook pro.
-
Yes the data is converted to sRGB.
-
Yes the data is converted to sRGB.
So why would I ever want to export as sRGB. It's like clipping yourself unnecessarily, couldn't you just leave it to the viewer to clip to their each respective display?
-
So why would I ever want to export as sRGB. It's like clipping yourself unnecessarily, couldn't you just leave it to the viewer to clip to their each respective display?
Because in this usage, sRGB is the correct output color space. Just as you would have to use a certain output color space for a printer.
-
There are two reasons you might want to convert to sRGB when exporting:
1) You want to control how out-of-gamut colors are clipped on an sRGB display, rather than leaving it to the viewer.
2) You do not trust the viewer to be part of a color managed system that will understand your embedded profile.
If you export an image in AdobeRGB, but the viewer ignores your profile and assumes sRGB, your colors will all be displayed wrong, even the colors which are within the sRGB gamut and are not clipped.
-
There are two reasons you might want to convert to sRGB when exporting:
1) You want to control how out-of-gamut colors are clipped on an sRGB display, rather than leaving it to the viewer.
2) You do not trust the viewer to be part of a color managed system that will understand your embedded profile.
If you export an image in AdobeRGB, but the viewer ignores your profile and assumes sRGB, your colors will all be displayed wrong, even the colors which are within the sRGB gamut and are not clipped.
I understand point (1). That's exactly what I'm trying to achieve. Control how the image will be viewed in sRGB space. But what I have found is that converting it to SRGB automatically with LR yields much better result than me manually going to into soft proof mode and tweaking the colors. Do you know why that is?
-
I understand point (1). That's exactly what I'm trying to achieve. Control how the image will be viewed in sRGB space. But what I have found is that converting it to SRGB automatically with LR yields much better result than me manually going to into soft proof mode and tweaking the colors. Do you know why that is?
In the latter case could you explain exactly what you do. Do you first soft proof using the sRGB profile and then export the resulting virtual copy to JPG in the sRGB color space?
-
In the latter case could you explain exactly what you do. Do you first soft proof using the sRGB profile and then export the resulting virtual copy to JPG in the sRGB color space?
That's exactly what I did. I opened the image in LR, in soft proof mode with sRGB with the destination profile. And then turn down the saturation of the yellow color on a flower until the clipping indicator goes away.
-
That's exactly what I did. I opened the image in LR, in soft proof mode with sRGB with the destination profile. And then turn down the saturation of the yellow color on a flower until the clipping indicator goes away.
But did you during that process manage to get the proof copy to look the same (or closely the same) as the original without soft proofing in Lightroom?
-
until the clipping indicator goes away
The clipping indicators aren't accurate. You'll get a much better idea by studying the histogram; any one or two channels clipping at either end is gamut clipping. And you also get a pretty good impression of how much.
-
Also note that if you have out of gamut yellows, turning down yellow saturation wholesale will also desaturate yellows that are actually in gamut. You need to be selective.
-
Also note that if you have out of gamut yellows, turning down yellow saturation wholesale will also desaturate yellows that are actually in gamut. You need to be selective.
Maybe that's the problem, how do I only work on the out of gamut colors?
-
Maybe that's the problem, how do I only work on the out of gamut colors?
By and large, let the profile handle this. There is only a minor degree of tweaking you can do anyway. Out of gamut (OOG) colors are just a fact of life and considering the only rational reason to convert to sRGB is to upload to the internet or screen presentation outside color managed app's, how much work do you want to go through?
See:
http://digitaldog.net/files/LR4_softproof2.mov
-
By and large, let the profile handle this. There is only a minor degree of tweaking you can do anyway. Out of gamut (OOG) colors are just a fact of life and considering the only rational reason to convert to sRGB is to upload to the internet or screen presentation outside color managed app's, how much work do you want to go through?
See:
http://digitaldog.net/files/LR4_softproof2.mov
Hey off topic...but in your video you said the monitor gamut warning is comparing between the monitor's gamut with the working space gamut. Are you sure that is the case? Because if that was true, how come when I switch the destination profile from one to another, the monitor gamut warning areas change? Wouldn't it always be the same?
-
Because if that was true, how come when I switch the destination profile from one to another, the monitor gamut warning areas change? Wouldn't it always be the same?
It should change, you updated the profile of which the display gamut overlay is calculated.
-
Andrew, yalag is actually pointing out a discrepancy in your video that I myself noticed a few months back.
In the video, you state that the monitor clipping indicator is NOT changing with the target soft-proof space and you even comment how this then is of questionable use. The current version of LR actually DOES change the monitor clipping depending on the target profile.
I think that one of two things happened when you did the video:
- there was a bug that didn't permit the monitor clipping to display correctly (you were using the beta, so somewhat understandable).
- the LR image display just failed to update when you were doing the video or it was incredibly slow in updating. This is actually something that can occur, even with the latest version of LR. I have noticed several times that the soft-proofing clipping indicators (both monitor and target profile) can update extremely slowly or sometimes not at all when making certain changes. I have seen this happen on several occasions when changing the camera profile, for example. Exiting and re-entering soft-proof mode will usually clear things up.
-
- there was a bug that didn't permit the monitor clipping to display correctly (you were using the beta, so somewhat understandable).
- the LR image display just failed to update when you were doing the video or it was incredibly slow in updating. This is actually something that can occur, even with the latest version of LR.
Yes, it was the beta version which was used for the video and then fixed. I haven't seen the failure to update on my end, but there's a comment on the YouTube video where someone else says that's something he's seen. The OOG overlay should update.
-
I haven't seen the failure to update on my end, but there's a comment on the YouTube video where someone else says that's something he's seen. The OOG overlay should update.
Yes, that was me. ;D
And as I said in that comment, I think the degree that the rendering intent will handle things depends on the image. I'm including the following in case it might be useful for yalag. I know that you already know all this, Andrew.
First, as you have said many times the clipping overlays don't tell you how far out of gamut the colors are so that can be a big deal. If the colors are only slightly out of gamut the small adjustment made, even with relative colorimetric, will be nearly invisible.
Second, even if the colors are significantly out of gamut it may not be that noticeable if those clipped colors are providing little in the way of detail information or if the detail provided by those clipped colors is very small. There will be some degree of hue shift but there will be little or no noticeable loss of detail.
On the other hand, if the clipped colors ARE providing detail information (variation in tones over the clipped areas), then relying solely on the rendering intent can be problematic. You can often see this in images of brightly colored flowers, where the colors are significantly out of gamut and those clipped colors cause a very noticeable loss of detail in the image.
For Lightroom users, I have found two ways that often help with OOG situations beyond just using the HSL sliders. One is easy; one is a little harder.
The easy way is to just use the Tone Curve while in soft-proof mode and pull down on the white point. This won't necessarily bring all colors into gamut but it will often resolve a large percentage of the OOG colors in photos of flowers and such - particularly when going to something like sRGB. It darkens the image some obviously but for on-line viewing in sRGB I find that it often works pretty well. The eye adapts to the lower brightness pretty well and you don't get the faded color that HSL adjustments will sometimes give you. It also seems to work better than the Exposure, Highlights, and Whites controls where LR tries to be smarter than you want it to be in this case.
The more difficult method is to use Adobe's DNG profile editor to make adjustments to heavily saturated colors in a custom camera profile. I still don't have one "perfect" but I have done some experiments and they work pretty good for non-commercial work. I think of it as a way to create my own "perceptual" rendering intent and by doing it with a custom camera profile, it works even for a v2 sRGB target.
-
Yes, that was me. ;D
And as I said in that comment, I think the degree that the rendering intent will handle things depends on the image. I'm including the following in case it might be useful for yalag. I know that you already know all this, Andrew.
First, as you have said many times the clipping overlays don't tell you how far out of gamut the colors are so that can be a big deal. If the colors are only slightly out of gamut the small adjustment made, even with relative colorimetric, will be nearly invisible.
Second, even if the colors are significantly out of gamut it may not be that noticeable if those clipped colors are providing little in the way of detail information or if the detail provided by those clipped colors is very small. There will be some degree of hue shift but there will be little or no noticeable loss of detail.
On the other hand, if the clipped colors ARE providing detail information (variation in tones over the clipped areas), then relying solely on the rendering intent can be problematic. You can often see this in images of brightly colored flowers, where the colors are significantly out of gamut and those clipped colors cause a very noticeable loss of detail in the image.
For Lightroom users, I have found two ways that often help with OOG situations beyond just using the HSL sliders. One is easy; one is a little harder.
The easy way is to just use the Tone Curve while in soft-proof mode and pull down on the white point. This won't necessarily bring all colors into gamut but it will often resolve a large percentage of the OOG colors in photos of flowers and such - particularly when going to something like sRGB. It darkens the image some obviously but for on-line viewing in sRGB I find that it often works pretty well. The eye adapts to the lower brightness pretty well and you don't get the faded color that HSL adjustments will sometimes give you. It also seems to work better than the Exposure, Highlights, and Whites controls where LR tries to be smarter than you want it to be in this case.
The more difficult method is to use Adobe's DNG profile editor to make adjustments to heavily saturated colors in a custom camera profile. I still don't have one "perfect" but I have done some experiments and they work pretty good for non-commercial work. I think of it as a way to create my own "perceptual" rendering intent and by doing it with a custom camera profile, it works even for a v2 sRGB target.
Thanks that was very helpful. I'll probably try the easy method, since the hard method is out of my expertise at the moment.
But back to my original question, as pointed out by another poster and Andrew's video. It's not even advisable to try to fix out of gamut colors to begin with. As demonstrated, letting the software do the clipping automatically often produces better results (unless like you where you have some very sophisticated method).
So then this begs the question, doesn't this apply to printing as well? When I soft proof my printer+paper profile in LR. What exactly is my goal here? I understand that it's useful to correct contrast etc. But when it comes to color, I should pretty much just toggle the intent and let it be. Any attempt to correct the color will probably yield a worse result than letting the software does its thing.
This whole experiment started when my objective was to export images for internet viewing (srgb). But the more I explored it here, I realize it's the same thing with printing.
-
But back to my original question, as pointed out by another poster and Andrew's video. It's not even advisable to try to fix out of gamut colors to begin with. As demonstrated, letting the software do the clipping automatically often produces better results (unless like you where you have some very sophisticated method).
So then this begs the question, doesn't this apply to printing as well? When I soft proof my printer+paper profile in LR. What exactly is my goal here? I understand that it's useful to correct contrast etc. But when it comes to color, I should pretty much just toggle the intent and let it be. Any attempt to correct the color will probably yield a worse result than letting the software does its thing.
Note that you are saying "OFTEN produces better results" and "PROBABLY yield a worse result". So one question to consider is what probability do those qualifiers represent. Is it >50% of the time, >80%, >95%, or >99.9%? Personally, I would have no problem believing the >50% and probably not much of a problem with the >80% either but I would start questioning probabilities much higher than that. I see plenty of pics on the web of red, yellow, and orange objects where the photographer just let the rendering intent do its thing and they look pretty bad.
The other variables that come into play are how perfect you need your image to be and how much time and money you are willing to spend to get it there.
So ultimately I think you need to be your own judge and to use your eyes to the extent that you can, with soft-proofing, test prints, and so on. Whether you are using relative or perceptual rendering intents, or you are doing your own tweaking to bring colors into gamut, you want to be looking for hue shifts and detail loss and making sure those compromises, if any, are acceptable to you given your time and money constraints.
-
Thanks that was very helpful. I'll probably try the easy method, since the hard method is out of my expertise at the moment.
Here's the deal. Ask yourself why you're moving from a nice wide gamut image to sRGB. Probably to post the image to the internet. Short of being forced to send to some lame lab that is cluless about color management, it is for web or screen viewing. If for the web, all bets are off in terms of what other's see of the image. Are their browsers color managed? Is the display calibrated? Is it a nice new Eizo or NEC or a 16 year old CRT? It's a crap shoot. While I believe we should make our images look as good as possible, is it worth the time and effort to muck around for conversion to sRGB to be posted to the web? Only you can answer that.
Having to resort to showing images on the web and expecting everyone see the same quality you do is like being stuck hungry in an airport with a 5 hour layover and only have the choice of KFC or McDonald's.
-
Note that you are saying "OFTEN produces better results" and "PROBABLY yield a worse result". So one question to consider is what probability do those qualifiers represent. Is it >50% of the time, >80%, >95%, or >99.9%? Personally, I would have no problem believing the >50% and probably not much of a problem with the >80% either but I would start questioning probabilities much higher than that. I see plenty of pics on the web of red, yellow, and orange objects where the photographer just let the rendering intent do its thing and they look pretty bad.
The other variables that come into play are how perfect you need your image to be and how much time and money you are willing to spend to get it there.
So ultimately I think you need to be your own judge and to use your eyes to the extent that you can, with soft-proofing, test prints, and so on. Whether you are using relative or perceptual rendering intents, or you are doing your own tweaking to bring colors into gamut, you want to be looking for hue shifts and detail loss and making sure those compromises, if any, are acceptable to you given your time and money constraints.
I'm just looking at it from an theoretical point of view as it's about education at this point more than practical applications. In practical terms, I would imagine it probably depends on a lot of things. The image itself, how far off is it out of gamut etc.
Theoretically, almost every course I have watched now recommends some kind of workflow where you would edit the photo during soft proof. But if >50% of the time, the editing puts you in a worse position, I think this should be warned or explained. It's not even an edge case.
One of these courses is Andrew's own color management with photoshop cc. I see that in his video, he carefully avoided the saturation adjustments and stuck with contrast and hue only. While many other courses recommend that you tone down the saturation. In fact adobe's own video (http://help.adobe.com/en_US/lightroom/using/WS2bacbdf8d487e582-5591e4a41341ae6cc6d-8000.html) does this! I really don't see this being a good idea unless you have a very specific image where pulling down the saturation of that color only have a very localized effect. Otherwise you'll almost always end up with a crappy image.
I guess one day if they come up with color control only for out of gamut colors then that would be a good workflow?
-
[...]I guess one day if they come up with color control only for out of gamut colors then that would be a good workflow?
I think that's a bit too simple of a solution. The best solution will vary based on the image itself, and the goals of producing the output. We have two common rendering intents that handle out of gamut colours in different ways: perceptual compresses the whole gamut to preserve subtle details at the expense of lesser overall accuracy of colour; while relative colorimetric preserves overall accuracy at the expense of clipping out of gamut colours thus wiping out detail. Sometimes one approach is okay, sometimes the other is better. It depends.
In the same way, if you chose to do manual work in post to adjust the image prior to letting the rendering intent tackle things, sometimes you might want to bring down only the out of gamut colours. Other times you might want to bring down a broader range of colours including some that are in gamut, in order to preserve subtle surrounding details and "feather" the adjustment more gracefully.
When I do these kinds of edits, saturation usually is the last adjustment I resort to. I usually start with tonality adjustments in one form or another -- curves, luminosity. Or other types of non-saturation adjustments like hue or density tweaks. The adjustments I make are guided by soft-proofing initially, but not by soft-proofing in Photoshop for cases I really care about. If I really want to see what's going on, I use Gamutvision (a Windows tool) which provides more comprehensive, detailed and useful information about differences between image colour and rendered output colour.
Whatever types of adjustments I'm using, I would rarely-to-never hit the entire image with a global adjustment. Typically I would make changes in selective regions (including specific channels) of the image, but covering a somewhat broader range of colours than just those that are strictly out of gamut. I'm usually trying to strike a balance between detail and colour accuracy; where I put the fulcrum for achieving that balance may vary from one image or output purpose to another.
And of course, like Andrew says, I don't go to this effort for any images I post at low res to the web. Those cases just aren't worth it. When I tackle this issue manually, it's for prints that I really want to dial in.