Am 04.07.2013 01:46, schrieb Graeme Gill:
Gerhard Fuernkranz wrote:Btw, I'm not sure if anybody has already noticed this yet, however, on a reflective IT8 target one can indeed find a subset of about 100 patches, which are almost (< 1 delta E deviation) "additive" (i.e. the reflectance spectra of these patches are (almost) linear combinations of three basis spectra).Hi Gerhard, isn't that a consequence of those patches being (effectively) composed of just three colorants ? - ie. that their spectra are a linear combination of three colorant spectra ?
Hi Graeme, I would not expect that. There are certainly just three colorants involved, but the mixing-model of the chemical printing process is not additive, but rather multiplicative (no half-tone raster involved, i.e. Neugebauer does not apply - rather Beer's law applies, but not exactly either). Maybe a non-linear PCA were able to represent the spectra of all patches with just three principal components, but it is is by far not possible to represent the _whole_ set of patches by _linear_ combinations of only three basis spectra with reasonable accuracy. Despite the non-linearity of the mixing process it is however still possible to find a 3-dimensional subspace of spectral space, in which as many as about 100 of the patches happen to lie... Best Regards, Gerhard