Forums

Light Illusion is renowned for the level of support it provides to all its customers, and strives to provide an unmatched level of assistance.

These forums are provided to enable open discussions on all aspects of the Film and TV industry.

 | Forums | Register | Reply | Search | Statistics |
Display Calibration Light Illusion Forums / Display Calibration /

Question about INPUT ICC color profiles

Author taranvh
Mono
#1 | Posted: 5 Dec 2018 21:02 
I've been tearing my hair out for the last month or so, trying to learn how computer color works, but the deeper I dig for answers, the more questions I have. Here's just one of my many, many questions...:

(Before I begin, please note that this question is NOT about a monitor's color profile, which I believe is called an OUTPUT color profile. I am asking about the color profile that is attached to an image, like a .jpeg, which I believe is called an INPUT color profile. Someone tell me if I'm wrong.)

On the Xrite Color Checker SG (pictured here: https://xritephoto.com/colorchecker-digital-sg)

...it looks like the second square is "supposed to be" 127,127,127. (In the sRGB color space)

However, imagine that when we take a picture of it, we get 125,128,122.

Let's say that we use that information to create an ("input?") ICC color profile, which is then... I guess, "uploaded?" onto the camera. Now the camera "has" that profile, and will "use" it for all its photos. But, exactly HOW does this work?

Let's say that we take another picture of the Xrite Color Checker SG, this time with our newly created ICC profile installed in our camera.

Question #1: Is an ICC profile just a lookup table, which says to change 125,128,122 into 127,127,127 (and so on?) Because it seems like that would result in a lot of doubled and skipped values in the final image.

Question #2: Is the ICC profile not immediately... applied? converted? baked? onto the image as it is created, so that the values are all now "correct?" That is, would the final photo would properly show that grey square as being 127,127,127?

Or is the profile just "included" with the photo, as about 4kb of metadata, and that square is still the incorrect 125,128,122, and the color profile must be referenced, in order to "correct" those values? Because, apparently some websites will remove that data, to save space, meaning that the photo is guaranteed to look wrong.

Thanks for reading/responding!

Taran

Author Steve
Scene
#2 | Posted: 5 Dec 2018 21:12 | Edited by: Steve 
I know of no camera that can have an ICC installed in it.
ICC's are used just within the computer domain.

The input ICC just defines the 'colour' from the input device, so the output ICC knows what to do with it so it 'looks' correct on the output device (display/printer).

And yes, ICCs are a minefield to navigate/understand.

That is why they are really not used in the professional film and TV industry.
We use 3D LUTs, as they are a lot more controllable and easier to understand/manage.
(Even though you can have 3D LUTs in v4 ICCs - but applying them is a nightmare to understand/get right.)

See: https://www.lightillusion.com/spaceman_manual.html

Oh, and the 'program' displaying the image with the ICC embedded in ti must be ICC aware.
Some are, some aren't - and that includes different browsers, and different graphics programs.

Steve

Author taranvh
Mono
#3 | Posted: 5 Dec 2018 23:53 | Edited by: taranvh 
Steve
ooooookay....
Then maybe you can explain what this means?

"An image, when it comes out of a camera, must contain the ICC profile of that camera to display the "right" colors (L*a*b*) but it is immediately and automatically converted into a neutral color space."
Source: https://www.color-management-guide.com/icc-profiles.html

That whole website is from a guy for whom English is his second language, which makes the whole thing hard to read. And I think there are a lot of inaccuracies.

He also later says this:
"Well Photoshop will read the matrix table of the ICC profile of this scanner and then it knows that when it has to display the RGB value 128,140,128 it must actually display the color L*a*b* 54,0, 0 (equivalent in RGB to 128,128,128) and it will then send a corrected RGB signal to the screen in style 128,116,128 to take into account the defects of the scanner in green in this example."

Which is why I'm asking about exactly how ICC profiles work... because that description sounds absurd.

Author Steve
Scene
#4 | Posted: 6 Dec 2018 08:48 
It means the individual that wrote the information doesn't understand ICCs...
You answered you own question with your last comment.

Steve

Author taranvh
Mono
#5 | Posted: 6 Dec 2018 18:41 | Edited by: taranvh 
Steve
Well, that's a relief.
But he's not entirely wrong. On the next page, he talks about assigning ICC profiles to images that you export from Photoshop.
I don't think you need to bother to read it, though:
https://www.color-management-guide.com/assign-icc-profile.html

My next question is about THIS article that talks about the same thing, but he has proof of the weirdness.
Go to this page, and scroll down until you get to the "Correct or wrong?" section:
http://regex.info/blog/photo-tech/color-spaces-page1
I dragged the middle picture onto my desktop, then opened it in Photoshop. I used the eyedropper to sample the sunglasses, and THIS was the result:
https://imgur.com/ZEYax9D
Now, obviously, those sunglasses appear BLUE, not RED, as the RGB values would imply. This is because of the "MySuperFunkyRGB.icc" profile that is ..."included" with the image. If you assign the sRGB IEC61966-2.1 profile instead, his skin turns green, and the sunglasses turn red. This matches up
with what the sampled sRGB values say.

Soooo, I'm wondering WHY it's even possible to have an image that has certain values, which are NOT the values that actually get displayed, when you apply an ICC color profile to the image. This surely serves some kind of use... obviously, this is an extreme example. But this whole concept doesn't make sense to me in any context. Why not just burn it in? Why have "incorrect" values, and depend upon a web browser to read a 4kb ICC profile (which has hopefully not been lost) just to properly display the image?

Author Steve
Scene
#6 | Posted: 6 Dec 2018 18:46 
The idea is to allow a wide gamut, or high dynamic range, image to be 'mapped' to output devices that do not posses the same capabilities.
But, understanding the concepts and workflows is not at all simple.
And maintaining any level of true accuracy is next to impossible.

Welcome to the world of ICCs...

Steve

Author taranvh
Mono
#7 | Posted: 6 Dec 2018 20:03 
Steve
But I still don't get it.
Let's say I've got an HDR photo. Not a faux-HDR photo, where the final image is already in sRGB... no, I'm talking like, a genuine 10-bit HDR image.
(Now, first of all, I don't even know if this is possible, on the web. Are there photos like this? All the "HDR" photos I can find are always already converted to sRGB.)
Or perhaps, the only way to see a "genuine" HDR photo, is to go to Netflix on a computer with an HDR screen, and load up something like Marvel's Daredevil, making sure you are watching the HDR version, and pause the video during a bright scene...
If you had Photoshop open, and you used the eyedropper tool to sample the color of that video... would it give you the answer in 10-bit or 8-bit? Would the eyedropper tool "know" that the image is HDR? Or would it treat it as sRGB? If it treats it as sRGB, does that mean it's using perceptual or relative colorimetric rendering intent, to do so??
I can't do this test myself, since we do not yet have HDR monitors at our office...

Oh, and this is important:
I thought that all devices would already have information about all the major color spaces already loaded onto themselves. So, all the most common color spaces within the categories of CIE, RGB, YUV, HSL/HSV, and CMYK. That way, the only metadata a photo would need would be: "This photo is adobeRGB. Display it in the best way you can," and the device would take care of the conversion into its own native color space, on its own.
But, I guess that's NOT how it works? Like, every device perhaps ONLY knows how to convert from a PCS (Profile connection space) like CIELAB into its own native color space? So, an ICC profile needs to have detailed instructions for how to convert the values of a picture's pixels, (which might be AdobeRGB) into a PCS (Let's say CIELAB), and only then will the device be able to convert from CIELAB, into sRGB, doing so according to information already contained on the device?
Because THAT would make sense. Please tell me I'm right...
Or, does the ICC profile ALSO contain information on how to convert from CIELAB, to sRGB? Because that seems crazy. It'd have to have info about how to convert to EVERY possible color space, because it doesn't know which one a device might be using. Surely it might only specify "please use relative colorimetric rendering intent, NOT perceptual." and that's all... riiiiight?

Author Steve
Scene
#8 | Posted: 6 Dec 2018 20:13 
As I say, welcome to the world of ICC profiles...
Not a lot more I can say.
You really need to research ICCs in-depth.
Unfortunately it's not something we can cover in a forum discussion.

But, you are confusing HDR video and HDR photography.
Not the same.
And as I say, ICCs are not used in the video world, especially home TVs, so there is no valid reference there.

And any capture device has its own colour space.
They 'may' be able to be limited to a smaller colour space, such as sRGB, but none do AdobeRGB, or anything like that.
They are all different.

Steve

Author taranvh
Mono
#9 | Posted: 6 Dec 2018 21:18 | Edited by: taranvh 
Steve:
You really need to research ICCs in-depth.

That's what I'm trying to do!
I've spent weeks on this. I've read literally hundreds of articles on this and other things related to video/computer color, and I am hardly any closer to understanding what the hell ICC profiles are supposed to be doing, because all they've ever done for me, is mess up Photoshop. Websites like the one I linked you to have forced me to backtrack multiple times, as I slowly realize that more and more of what I've already read is inaccurate or outright false. Not even Wikipedia is trustworthy - either having not enough info, not enough sources, linking to sources that are offline, or yes, getting some things wrong (for example, sometimes referring to CIELAB as a color space, sometimes referring to it as a color model... I still don't know which one is right, or if it's both...)

I'm only on this forum because I took a Lynda course from Robbie Carman, had lots of questions, and emailed the people over at Mixinglight.com, who suggested I ask my questions here.
So if not here, then where can I go to learn about ICC profiles?

I just want to know what "working space" to use in Photoshop and why,
Why DisplayCAL 3.7.1 has done SOMETHING to change the way my monitor looks, but it seems like it did NOT use a .ICC profile to do it, but rather, a .icm?
And why this CANNOT be set through Windows' own Color Management panel, but ONLY through DisplayCAL...

And a hundred other questions...

Author Steve
Scene
#10 | Posted: 6 Dec 2018 21:31 
ICM is basically just a Windows term for an ICC...
And I suspect the one you generated can only be used by Displaycal as Displaycal has overridden the Windows colour management operation.

We really can't help further as we are really here to support our customers.
Other forum members may be able to offer assistance, but as many are just as confused about ICCs, and their operation, I'm not sure there is much additional help available.

ICCs are just a real pain to try to use.
And as you are finding out, there is little real understanding anywhere...

Steve

Author taranvh
Mono
#11 | Posted: 6 Dec 2018 21:34 
Well, thank you for your help.
I'll continue elsewhere on my frustrating quest for answers...

Display Calibration Light Illusion Forums / Display Calibration / Question about INPUT ICC color profiles Top
Your Reply Click this icon to move up to the quoted message
 

 (?)
Only registered users are allowed to post here. Please, enter your username/password details upon posting a message, or register first.
 
Online now: Guests - 2
Members - 0
Most users ever online: 411 [17 Mar 2015 10:47]
Guests - 411 / Members - 0
 
© Light Illusion - All rights reserved - Privacy Notice