large gamut color space

[atvinnys]

I was just reading a post from Colin about color space (related to Firefox).
there, he mentioned that given the 8 bits encoding for each primary color, we got a max number of combination of 16 millions or so...

now, given that encoding is the same on any color space, if you have a large gamut color space (say ProPhoto), maybe you get more possible colors , but you are also spreading those values to reach those corners of the gamut...

so if you have a pic that has lots of variations around the same color, I would think that all color variants would not be encoded in a large Gamut color space, but could actually gain using a smaller gamut, bc there will be more possible encoded values into that small color variation spectrum...

did I miss anything?
Vincent

[Colin Southern]

did I miss anything?

Yes and no :)

I don't recall saying anything about 3x 256 bit channels adding up to the 16M colours etc (even though they do), although I did post about using 8 bit (256 colours per channel) as it applies to absolute levels in gamuts (re displaying in non-colour-managed browsers), which is probably what you're referring to.

now, given that encoding is the same on any color space, if you have a large gamut color space (say ProPhoto), maybe you get more possible colors , but you are also spreading those values to reach those corners of the gamut..."

Yes. Keep in mind also that "conceptually" the gamut of the human eyes is also very limited compared to what some colour spaces are capable of representing (in a kind of backward looking kind of way!).

so if you have a pic that has lots of variations around the same color, I would think that all color variants would not be encoded in a large Gamut color space, but could actually gain using a smaller gamut, bc there will be more possible encoded values into that small color variation spectrum..."

Excellent observation - and this is the reason one should never, ever, use 8 bit colour (256 bits per channel) on large spaces - you need to use 16 bit (of which photoshop actually uses 15 bits per channel) which gives you 32,756 possible steps per channel (up slightly from 256!) - I was only using the 8 bit representation to make the concept of why a large-gamut encoded image looks dull & unsaturated if the displaying application incorrectly assumes that it's sRGB.

There's a raging debate going on over 8-Bit -v- 16-bit manipulation, with little agreement - what little agreement that many do have however is that the potential does exist for image degradation for large changes in big spaces with 8 bit representation. Personally, I just do everything in 16 bit - it errs on the safe side, and I don't have size or performance issues on my processing PC. Printing often needs to be 8 bit for many, but since I do all of my own, this doesn't affect me either (I often just leave the images in LAB colour - my favourite working space).

Cheers,

Colin - pbase.com/cjsouthern

[atvinnys]

colin,

thanks for the answer...for some reason I got stuck with 8 bits....

[Colin Southern]

You're welcome,

The whole 8/16 bit thing and the many different colour spaces is something that trips up a lot of people - I've been guilty of having a quick print done in town (one not worth cranking my printer up for) - only to discover I've left it in ProPhoto and have to fix it and print it again, but unfortunately that's the way it works for now.

Cheers,

Colin - pbase.com/cjsouthern