Re: use of sRGB as a default
Re: use of sRGB as a default
- Subject: Re: use of sRGB as a default
- From: Graeme Gill <email@hidden>
- Date: Mon, 21 Jun 2004 15:43:37 +1000
- Organization: Color Technology Solutions
Matt Deatherage wrote:
OK, stop right there. Most discussion I've read (including Apple's
developer documentation) calls sRGB a _color space_, but you just
called it a "profile" and a "source profile." I thought a _profile_
defined a device's behavior. I'm not meaning to pick on you, because
there clearly is such a thing as a color space profile, but if people
use such key terms interchangeably when they're not, there's no way
anyone is ever going to figure anything out. I'm beginning to think
the unmodified word "profile" should never be used except to refer to
an ICC file on disk, and that in all other cases, it should be "device
profile," "display profile," "color space profile," etc.
Technically a device profile defines a colorspace. The fact that the ICC format
distinguishes between device profile colorspaces and non device colorspaces is
peculiar to that format. sRGB can equally be regarded as a device profile
(since it is intended to represent a generic display device behavior), or
as a pure RGB colorspace unrelated to any particular device.
The ICC spec. requires that a colorspace conversion profile be LUT based
(though it might contain matrix/shaper data as well), while a display
device profile must be matrix/shaper, which can be more compact and precise
in its conversion, making it (strangely) more suitable for defining a
standard RGB colorspace.
[A LUT profile could be configured to emulate a shaper/matrix profile,
but I'm not sure how commonly this is done.]
So, as a programmer, do I *ever* embed a display profile? If not, can there be
> a simple statement somewhere that says "display profiles are always destination
> profiles and never source profiles?" (I may have missed it, but if it's that simple,
> it *should* be that simply said.)
There are two reasonable models I can think of for dealing with user interaction with
an image on their display:
1) Direct. Convert the image into the display colorspace from whatever colorspace
it is currently in, let the user manipulate in display colorspace, write the image
directly to the display, tag the resulting image with the display profile, or
convert to some other colorspace and tag with that profile before saving.
(If the actual display colorspace is poorly defined or poorly behaved in terms
of gamut or quantization, this can lead to trouble.)
2) Indirect, the underlying image is kept in some working colorspace. Convert the image into
the working colorspace. Let the user manipulate the working colorspace values. Every
time the image gets displayed to the user, convert between the working space and
the display space (the OS/CMM may do this for you). Tag the resulting image with
the working space profile, or convert to some other colorspace and tag with that
profile before saving.
It's really not that complicated (ignoring gamut issues).
Graeme Gill.
_______________________________________________
colorsync-users mailing list | email@hidden
Help/Unsubscribe/Archives:
http://www.lists.apple.com/mailman/listinfo/colorsync-users
Do not post admin requests to the list. They will be ignored.