In order to properly calibrate an electronic camera for a variety of illuminantsit is necessary to estimate the spectral sensitivity of the camera. This spectral characterisation is obtained by measuring a set of samples of known spectral re ectances and by inverting the resulting system of linear equations. In the presence of noise, this system inversion is not straightforward. We describe several approaches to this problem. In particular we show that the choice of samples is of great importance for the quality of the characterisation, and we present an algorithm for the choice of a reduced number of samples.