RE: 8 bit vs 16 bit profiling -- does anyone gives a hoot?
RE: 8 bit vs 16 bit profiling -- does anyone gives a hoot?
- Subject: RE: 8 bit vs 16 bit profiling -- does anyone gives a hoot?
- From: <email@hidden>
- Date: Thu, 7 Jul 2005 10:21:44 +0200
- Organization: 8x10.se
It might be that the ACE uses high precision internally, but PS seems to be
optimized for speed in conversions to monitor and strips the source data
down to 8 bits before conversion if your scale is 50% or smaller. Try
converting your 16-bit source data to a linear profile and watch the shadows
block, then zoom in to 100% view and all the shadow detail comes back.
- Lars
> -----Original Message-----
> From:
> colorsync-users-bounces+colorsync=email@hidden
> [mailto:colorsync-users-bounces+colorsync=email@hidden.
> com] On Behalf Of Graeme Gill
> Sent: Thursday, July 07, 2005 06:17
> To: email@hidden
> Subject: Re: 8 bit vs 16 bit profiling -- does anyone gives a hoot?
>
> Roger Breton wrote:
> > So I heard, that ACE is internally implemented as 16 bit.
>
> I understand that it actually uses 32 bits or more internally.
>
> > You mean ACE could interpolate the LUTs in 16-bit instead of 8 bit?
>
> No, I'm pointing out that there is a difference between how the
> numbers that determine a transform are stored, and how they are
> represented during calculations. Of course an 8 bit ICC profile has
> sacrificed some precision in the name of smaller files, but that
> doesn't automatically imply that all input, output and intermediate
> calculations in the transform are 8 bit.
>
> > Gotta back to icclib for further testing of my theory, I guess. But
> > the problem I'm seeing is that InDesign, Illustrator and maybe
> > Photoshop too (I stand corrected) are all using 8-bit
> conversions. In
> > that context, I wonder what the value of using "Large" or 16-bit
> > profiles in the first place? When everything will be dumbed
> down to 8
> > bit with the possible artifacts I discussed above. Does an
> 8 bit profile by design offers less grid points?
> > Will result in the the CMM doing larger interpolations?
>
> I'd be kind of surprised if this was the case. Most of the time you
> are linking two profiles, even if you are only going to use the linked
> result to transform 8 bit files, so extra precision in the individual
> tables (input, clut, output, *2) will help prevent accumulating
> rounding errors at each stage.
>
> Apart from a minor point (the input & output tables are fixed at
> 256 entries, but remember entries are interpolated between), the
> actual clut resolution can be the same as 16 bit profiles.
> What is lost is the precision of the numbers in the tables.
>
> I have the impression that actually 16 bit profiles are the norm, and
> that 8 bit profiles are rather rare.
>
> Graeme Gill.
> _______________________________________________
> Do not post admin requests to the list. They will be ignored.
> Colorsync-users mailing list (email@hidden)
> Help/Unsubscribe/Update your Subscription:
> email@hidden
>
> This email sent to email@hidden
>
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden