Am 19. September 2015 15:01:18 MESZ, schrieb Wolthera <griffinval...@gmail.com>:
>I sent a mail back in august about retrieving data from an icc profile. >In >the case of Linearising a given colour value from a gamma corrected >space >for the use in filters, the following was suggested: > >1. Get a linear version of the same profile. >2. Convert with LCMS to that profile and back when you're done. > >Problem is, how does one obtain this linear profile when expecting any >arbitrary icc profile? Obtaining a linear version is indeed not practical. Working spaces are mostly matrix shaper profile. From there might come the assumtion that creating a linear version of such profile is almost trivial. But table based profiles are similarily possible - and then obtaining a linear version is not trivial. >There's something of an 'linearisation device >link' >mentioned in the API docs, but it doesn't mention how to use this, so I >can't verify if this is the function that I am looking for. Linearisaton profiles are created from per channel measurements to bring a press in good profiling condition. It is as well used to keep print conditions stable according ro measurements. So that is not exactly your case ;-) >The second option is to have LCMS customly make a profile from the data >of >the abitrary profile, but this seems very error-prone? You will need much effort to get that right. >Option 3 is to have several hard-coded rgb spaces, which is also quite >awkward. It would have its merits. You can then clearly define how a given filter will work and process in linear space without clipping >What would be the best way to solve this? I guess you need to discuss if *all* filters shall work in each space differently. For curve based filters I would assume yes, each space shall work differently for per channel curve operations (e.g. curve tool). For such tools you do not need to colour convert/work in linear space. For doing a desaturation effect users might simply expect a neutral desaturation regardless of the underlying colorspace. For blending effects, I would assume CIE*XYZ (with linear gamma, processed as floating point precission) will give most relyable/repeatable results. So, sorry, no one single answere. More a - it depends on the situation. kind regards Kai-Uwe ------------------------------------------------------------------------------ _______________________________________________ Lcms-user mailing list Lcms-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lcms-user