|
About three centuries ago Newton discovered that white light is a “mix of rainbow’s colors”. Since then scientists and artists tried to bound the concept of colours into some definition and into collections and catalogues of colours. The results was to have many theories and many ways of classifying colours. The study of human vision and the way our retina works suggested a tristimulus model that began a mathematical model in the ’30 of last century thanks to the work of the CIE (cie.co.au). Although the model evolved in its almost one century of life, the basis are the same of the original one. The model has a couple of huge advantages: it allows to assign to each colour some numeric colorimetric coordinates that can be propagated across any media to “communicate the colour”, and, being based on human physiology, it is able of predicting our perception of colours and colour differences. In the era of digital communication, practically speaking the 100% of images and colours are represented following this model and, if all the colour management chain is correct, we can hardly distinguish the original colour of a flower from its representation on a colour monitor. So one could think that the problem of representing colours is solved. Unfortunately not at all. Reducing a colour to three numbers (the colour coordinates, XYZ or Lab) is an irreversible process, so given XYZ coordinates these could correspond to infinite number of spectra (each of which would give to our eyes the same colour stimulus). This ambiguity generates the metamerism. The spectrum of colour of a flower will be not the same spectrum of the representation of that flower on a monitor, even if they seems to be the same colour. Digital colour imaging fidelity is totally based on the metamerism. We most that we can say of two colours is that they are colorimetrically equal. Colorimetry allows to communicate a stimulus that induces a specific colour in our eyes, it doesn’t communicate the real physical nature of that colour. Every time we need to associate colour to matter, colorimetry become useless. Only a complete spectrum can do that, and this brings us to the need of a spectral imaging approach in all the fields where matter matters. In the first part of this work the authors propose a flexible approach based on custom multibandpass filters and AI calibration that allows to achieve the needed spectral resolution with a tuning of the amount of spectral images used, that will be always less that the classical “one band – one filter” approach. In the second part the spectra obtained from images are compared to spectra measured with laboratory spectrophotometers to establish the reliability of the approach. The third part is a collection of spectral imaging case studies in the field of Cultural Heritage, Architecture and Archeology, where identification of matter was a key factor more than having digital images with colorimetric high fidelity.
|
