This is indeed one of the best explanations of color management I’ve read so far (and I’ve read several ….). Congratulations and thank you !!

One question I have:

On your sample images, everything works as expected in browsers (no Color Management in IE7, Color Management in Firefox 3.5) as well as in imaging applications (IrfanView, LR, PSE).

This is true for all color codings except “Jeff’s LCD”. Looking at the images with “Jeff’s LCD” color profile embedded, I notice a slight shift in exposure/tint, especially well visible on the green cup in the middle of the image.

I imported 2 images(ProPhoto and Jeff’s LCD) into LR, which shows the same difference. Again, all images of a certain subject look exactly the same, except the one with Jeff’s LCD.

To illustrate what I’m talking about, I have exported the two (different looking) images from LR encoded in sRGB, they can be downloaded from https://www.yousendit.com/download/Z01PL0dEVEhsamV4dnc9PQ.

I’d be very interested to know what’s going on here.

Thank you
Beat Gossweiler
Switzerland

I finally discovered the answer to my question–and now realize that I asked my question in a way that made it sound like I would only be satisfied with a 100% perfect translation from my screen to the viewer’s screen, which is not what I was really meaning to ask.

What I was looking for is Photoshop’s “Edit ->Convert to Profile” command. Assuming I’ve got a properly calibrated monitor, I can use Convert to Profile to translate my image from the ProPhoto color space that my DNGs start off in and into the sRGB color space that generally looks the best for most people’s uncolor-managed browsers / uncalibrated computer screens. (So they go from ProPhoto DNG => ProPhoto JPG => sRGB JPG)

I can also use the Convert to Profile command to take that original master image and convert it into the ICC profile of my local Costco Photo Center’s Fuji Frontier printer (as profiled by Dry Creek Photo), so that my prints end up looking virtually identical to what I’m seeing on my computer screen. Those printers ignore ICC profiles, so to get them to print correctly you must speak to them in precisely their own language (i.e. color space) as defined by their custom ICC profile (hopefully) listed on Dry Creek Photo’s site. To do that you use Convert to Profile with the destination profile set as the printer’s profile, but you don’t need to bother embedding the color profile because it’s now moot for two reasons: (1) the printer ignores it anyway and (2) even if the printer didn’t ignore it it wouldn’t do anything differently with your image because you’ve just fed it an image that is in precisely the color space the printer lives in.

Hopefully that helps people out there who are having the same problem I was.

Thanks again for all the great info, Jeffrey.

Thanks for the amazing write-up about color-spaces. You’ve done a phenomenal job. It answered a ton of my questions.

But I do have one question whose answer has alluded me for months, and I believe you must know the answer because you would have had to in order to create the example images at the top of this page.

So here goes: Other than the soft-proofing techniques discussed here and here–which require you to duplicate and manually adjust the colors of your web-destined image in Photoshop’s proofing mode while comparing them side-by-side with your originals–how can you take an image from, say, the AdobeRGB color-space and convert it to the sRGB color-space in such a way that it will display in precisely the same way regardless of whether the viewer’s web browser is color-managed or not?

Is there no automated & precise process for doing this without having to painstakingly eye-ball each image that I want to make “web proof”? … i.e. make them look the same for a given user when they view the image in Safari & IE side-by-side?

There is no mystical, Utopian “web proof” color space. The best you can do is work with a well-calibrated, well-profiled color-managed system so that those who consume your images have at least a chance to see the proper colors, if they, too, have a well-calibrated, well-profiled color-managed system. Everything else is just random. Ensuring that you produce sRGB (as opposed to, say, ProPhotoRGB or AdobeRGB) helps decrease the randomness for some consumers, but it’s all still random. Someone using a browser that’s not color managed probably doesn’t have a well-calibrated, well-profiled system anyway. Those who happen to have a color-managed browser likely don’t have a well-calibrated, well-profiled monitor anyway. So there’s much randomness, and it’s my opinion that any manual tweaking you do to “fix” a correctly-profiled image so that a user with a broken system might see it better — two wrongs canceling themselves out, you hope — is destined to fail horribly in all but the very specific cases you check against (e.g. you might end up making it look just dandy on your wife’s generic PC), at the cost of merely stirring up the randomness for most others, and explicitly breaking it for those few with good systems. —Jeffrey

It’s been weighing on my mind to update for that since FF3 came out in beta. Your comment prompted me to finally make some updates, thanks. —Jeffrey