Re: multichannel profiles
Re: multichannel profiles
- Subject: Re: multichannel profiles
- From: Phil Green <email@hidden>
- Date: Thu, 16 Nov 2000 17:12:46 +0000
- Organization: LCP
"Isaacs, Jeff" wrote:
>
A quadrillion, but I'd hate to be the guy who has to read the patches.
Or print them - you'd need about 1x10^12 sheets of paper! But it doesn't
seem necessary to me to have 1x10^15 possible colorant combinations when
you won't distinguish more than around 2x10^6 separate colours.
Unless you are using additional colorants to improve metameric matching,
the main reason for using >3 primaries in a reproduction is to increase
chroma. For any new primary, combining with any colour other than one of
the two adjacent ones in the hue circle will reduce chroma, giving rise
to a colour that you could have matched with the adjacent primaries +
black. So if you're only mixing adjacent colours, the possible number of
combinations is 4*n, where n is the number of primaries, multiplied by
the number of steps you want to have within each colorant combination
(at 1% dot intervals, this is 1x10^5).
The main weakness of computing n-colour separations from CMYK is that
you've already compressed the gamut to that of the CMYK media. When you
expand the gamut by adding extra colorants, you're then applying the
gamut increase uniformly, instead of selectively to the colours of the
original that require it. If you had a mapping directly from a CIE
colour to the n-ink set, you could get a better match to the original.
However, you could also achieve this by working with a reference
large-gamut CMYK space, thus avoiding the complexity of the CIE->n-ink
transform.
>
joe borne wrote:
>
> So your initial
>
> necessary number of patches jumps to 1,000,000,000,000,000.
--
Phil Green
Colour Imaging Group
School of Printing and Publishing
LCP
Elephant and Castle, London SE1 6SB
Tel: +44 020 7514 6759 Fax: +44 020 7514 6772
http://twinpentium.lcp.linst.ac.uk/digitalcolor/cig