first cs depend case: brighness,contrast

[Boudewijn Rempt]

On Tuesday 07 June 2005 09:00, Casper Boemann wrote:

> THIS by the way is a good time to ask: do we want say brightness filter  to
> produce the exact same visual effect independent of colorspace or is
> approximation good enough.

I think that doing what seems best in the context of the particular colorspace 
is the best option; it's not necessary to havce the exact same results with 
the color filters in different colorspaces. I think...

> I'm not talking of letting the effects be totally wrong, but more like the
> artist choosing to do it in CMYK,LAB or whatever because it produces a
> slightly different effect.

Exactly...

> Personally I think that say brighness filter should produce the exact same
> visual effect in every colorspace.
> doing:
> RGB >> change brightness+5
> should equal exactly (except for rounding errors)
> RGB >> CMYK >> change brightness+5 >> RGB
>
> If it is even possible without taking profiles into consideration is
> questionable, but something to strive for I think.

Well, I think it'll be difficult. Casper -- are you reading the lcms mailing 
list? With the kind of work you're embarking upon now it could be useful. 
Many problems have already been solved, or can be solved using the more 
esoteric lcms functions, like:

_________________________________________________________________________________

cmsHPROFILE cmsCreateLinearizationDeviceLink(icColorSpaceSignature ColorSpace,
                                            LPGAMMATABLE TransferFunctions[]);
_________________________________________________________________________________

 Creates a specialized devicelink, operating in space "ColorSpace". This 
devicelink
 has only linearization tables. Usefull for embedding 
gamma/brightness/contrast controls
 in a multiprofile transform.


  TransferFunctions[] must contain as many curves as components has the target
  color space. For example, RGB must be 3 curves. CMYK needs 4 curves.

-------------

Which to me reads as if lcms has methods to change curves in a colorspace 
independent fashio already. For the Krita colorspaces that are represented in 
lcms, we should strive to use these functions.


-- 
Boudewijn Rempt 
http://www.valdyas.org/fading/index.cgi
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://mail.kde.org/pipermail/kimageshop/attachments/20050607/bd312331/attachment-0001.pgp