After a lot of reading I could answer this myself: CATs, even the older ones are based on correlated color experiments and does model the color inconstancy aspects of human vision. Thus it's not correct to use them for "relighting", but people do anyway as there are no other established standard models for doing so. The linear Bradford seems to be slightly better than CAT02 for relighting applications, while CAT02 is the current best to model color appearance.
So if you make a camera profile for StdA and have reflectance spectra available you should first calculate StdA XYZ values from spectra, and then use CAT02 to get to D50 required for the profile connection space. This way the slight appearance differences are kept, within the accuracy limits of CATs, which is 3-4 DE for a StdA to D65 transform. If you want to model 100% color constancy (not true to human vision, but may be good for copy applications), also calculate the D50 XYZ values from spectra. If you need a "relighting transform" (that is want to predict the XYZ values for an object in a new light, rather than predicting a corresponding color which CATs do), look for example at this paper: http://www.cs.sfu.ca/~funt/2014_CIC22_MIRZAEI_GaussianPrediction.pdf /Anders On 05/28/2015 01:07:48 PM, Anders Torger wrote: > I've the author of DCamProf an open-source camera profiling tool that > can do DCP and ICC profiles. I use LCMS2 for CAT02 and various other > color management stuff. > > Recently a question came up regarding how calibration in light > different from D50 should be handled. Let's use StdA in this example. > > Currently DCamProf does like this: as it knows the spectra of the > illuminants and the test patches (and observer CMF) it can regenerate > XYZ reference values from scratch using spectral integration. This > means that regardless if you calibrate for StdA or D50 the XYZ > reference values will be calculated for D50, as the PCS is in D50. > > This means that the software makes profiles that try to make colors > look like they were shot under D50. In other words an artwork shot > under StdA and processed with an StdA profile will (ideally) produce > the exact same result as if shot under D50 and processed with a D50 > profile. > > Then someone told me that I should instead in the StdA case make XYZ > reference values for StdA and then convert to D50 (like the PCS > require) via CAT instead of regenerating from spectra. The reason > would > be that the CAT takes into account the minor appearance differences > that is between D50 and StdA viewing condition so those will be > preserved. > > I know there are such appearance differences in reality (it becomes > obvious in more extreme light temperatures) but prior to this my > humble > understanding was that a CAT, including CAT02 (with average viewing > condition in/out), did not intend to model those appearance > differences > but instead just tries to make a result as equal as possible to > "relighting" (ie recalculate our D50 values from spectra in our > case), > > and at the same time make some mathematical tradeoffs to make the > transform reversible. That is a CAT result is just an approximation > of > > the relighting result and any differences from actual relighting is > not > due to it tries to model slight appearance differences, but just > simply > limits in how precise the CAT can be. > > That is if my understanding is right it's always better to regenerate > from spectra if you have spectral information and only use CAT if you > can't. > > So who is right? > > /Anders > ------------------------------------------------------------------------------ > _______________________________________________ > Lcms-user mailing list > Lcms-user@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/lcms-user > > ------------------------------------------------------------------------------ _______________________________________________ Lcms-user mailing list Lcms-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lcms-user
