benifits of a linear color space?

[Zilly]

Ok what are the benefits of working in a linear color space ???

thank you :D

[tim]

I've read an entire book on color and I don't know what you're talking about.

[Damo77]

I assume you're referring to the "working spaces" (Adobe's terminology, I believe), ie sRGB, Adobe RGB etc.

They are "linear" (or you might say "perfect") colour spaces because - loosely speaking - they have smooth gamma curves from 0 to 255.

("Curve" and "linear" aren't the same thing, of course. To my understanding, a truly linear gamma is 1.0, but of course the common working spaces are 2.2.)

The working spaces are theoretical spaces - no real device has perfect "linearity", be it a scanner, camera, monitor, printer, etc. If they did, colour management would be a breeze!

Before the days of modern colour management, we did work in these non-linear spaces. But they're a dog to work with. The modern methods of working in one of these theoretic "perfect" spaces is much easier.

At the point of raw processing, we convert the camera data to a working space, and it stays there until print time.

I've just read what I've written, and it might sound like I have a clue. I don't. I've just read a lot of books, and this is my understanding. I hope someone can do better.

[gcogger]

I believe the OP is talking about using a working colour space with gamma = 1. This is the favourite idea of a guy on this website:

Timo Autiokari's pages (http://www.aim-dtp.net/index.htm)

There's a lot of info there which explains why Timo thinks that working in a linear space is a good thing. AFAIK, however, the vast majority of the industry disagrees with him :)

[tzalman]

Camera sensors record light levels (luminousity) linearly (o.k., Damo, nearly linearly) - each doubling of light intensity per unit of time produces twice the reaction from the sensor - and human vision is not linear - we are more sensitive to lower (darker) light levels and less so to bright light. It's an evolutionary self-defense thing. After demosaicing the digital image is still linear (gamma 1.0) and the next step in the chain is WB. That is because the action required to balance the data is itself linear. If in order to achieve WB the red channel needs to be strengthened by a factor of two, the data from the shadows needs to be multiplied by 2, the midtones by 2 and the highlights by 2. Since the action is linear it makes sense to do it when the image data is linear; the mathematics are so much simpler. Some converters also do NR at this stage, before the noise structure is contorted by the next step.

However, a linear image appears very dark to a human observer. In order to put the tonal values into the same relationship as human vision, the shadows need to be expanded and the highlights compressed by the application of a gamma 2.2 curve. At the same time, by applying separate curves for each channel, compensations can be made for the variations from strict linearity alluded to by Damo, by not sticking strictly to the 2.2 curve. IOW, the camera profile is combined with the gamma correction. We now have an image that more or less conforms to the way we see the world and we can use it as a basis for comparison with our memory of the scene to decide how it needs to be further adjusted.

So the short answer is that linear is good for the setting of WB and perhaps some NR, but not for anything more.

[Damo77]

Wow! Great info - thanks Tzalman.

[tim]

This could be an interesting thread. I better go read that book again.

[Zilly]

So the short answer is that linear is good for the setting of WB and perhaps some NR, but not for anything more.

Would color correction also be worth while doing in a linear color space

[tim]

Would color correction also be worth while doing in a linear color space

What problem are you having with the current solution to WB? I could send a message to Tom Knoll if you're having issues ;)

[Zilly]

Im not having issues with wb and photography in genral.
Im currently working on my Final major prodject for my media moving image course (its a 7 week final exam).
Im currently working in 32 bpc float to use the CGI units and plugins to maximum effect. (Still with me so far)
Now from what I understand when working in 32bcp float its advisable to work with a linear color space. Part of criteria for the grading on my Final major prodject is that you give reasoning for why you did what you did. So Im currently trying to research the benifits of working in a linear color space at the moment with out much luck :(

[Az2Africa]

I think the subject is worth reading up on. I'm going to take a look.

[gkuenning]

Im not having issues with wb and photography in genral.
Im currently working on my Final major prodject for my media moving image course (its a 7 week final exam).
Im currently working in 32 bpc float to use the CGI units and plugins to maximum effect. (Still with me so far)
Now from what I understand when working in 32bcp float its advisable to work with a linear color space. Part of criteria for the grading on my Final major prodject is that you give reasoning for why you did what you did. So Im currently trying to research the benifits of working in a linear color space at the moment with out much luck :(

I take it that by "32bpc float" you mean using a 32-bit floating-point number for each color (i.e., bpc = bits per color). If so, using a linear versus a nonlinear color space isn't likely to make any measurable perceptual difference as long as you're consistent and as long as you convert to an appropriate output space when rendering final images.

The reason is that 32-bit floating point has much more range and precision than any output device. You wind up with roughly 24 bits of precision per color on each pixel, or about 1 part in 16 million. You also have a range of about 10 to the +/- 38th, or 76 orders of magnitude. That means that you can represent lighting variations inside the deepest unlit cave to within .000006%. You can also represent the light on the surface of the sun at the same accuracy level.

Both are a wee bit better than my monitor can do...

[Damo77]

To bring this back down to my level ... your image would look TERRIBLE on the web! It would make the old sRGB vs Adobe RGB problem look insignificant ;)

[gcogger]

Can you imagine all the threads "My images look different in my browser compared to Photoshop"?

:)

[Damo77]

It would make everyone install Safari pretty quickly ... then the other browsers would have to catch up in a hurry. So perhaps it wouldn't be a bad thing after all.

[nadtz]

It would make everyone install Safari pretty quickly ...

Too bad safari sucks =( I'm still waiting for color management in firefox.