Re: [Lcms-user] Softproofing with parametric tonecurve

[Tobias E. <me...@ho...>]

On Donnerstag, 23. Juni 2016 19:13:26 CEST Marti Maria wrote:
> Ok, I see. A profile created in this way, with primaries and parametric
> curves, behaves like a "perfect" profile. That is, it can be inverted with
> no loss and its gamut is infinite. You can operate this profile with
> negative rgb or xyz values. Or with values well over 255. And the roundtrip
> is exact. So, if you tries to emulate which artifacts the profile causes,
> the answer is none and this is what you are getting in your softproofing.
> Built-in srgb behaves in same way.
> 
> Another thing is when you flush the profile to disk. Then you are forcing a
> quantized representation (in the case of v2)  Also, doing that imposes a
> gamut because no negative numbers or highlights over L* 100 are allowed in
> the file format. Then softproofing shows you the limitations the file
> format has.
> 
> So you have to choose whatever the ideal representation and the file based.
> 
> Hope this makes sense to you

Thanks, I changed my code to always take a roundtrip through dumping/reading 
for softproof profiles and it seems to work fine. However, I am a little puzzled 
by that design choice. I guess (just a gut feeling) that most people would use 
softproofing to see the result they get on a physical medium – or anything else 
that clips the colors. Especially when specifying a rendering intent for the 
transform.

Would it be possible to add a transform flag like cmsFLAGS_SOFTPROOFING_CLIPPED 
that does a clipping in an appropriate place? That might be less surprising 
and avoid the overhead of creating a new profile on the fly.

And maybe document the general issue where softproofing is explained?

> Regards
> Marti

Thank you a lot, even if you don't consider adding what I proposed. lcms is an 
awesome help and I wouldn't like to do without it.

Tobias