In this work we evaluate the performance of a multispectral system for the estimation of spectral reflectance curves associated to the iris of human eye. The aim of the study is to check the feasibility and the accuracy of the developed system in order to perform this kind of measurements, and to optimize the corresponding experimental set-up. In the experiment, twenty-six different subjects with different iris coloration were measured. We tested several sets of filters to be used as channels of the acquisition system, which consisted of a 12-bit depth cooled CCD monochrome camera. We also analyzed different reconstruction algorithms, specifically the pseudo-inverse technique and the principal component analysis, and several metrics such as the Root Mean Square Error and color difference equations were used in order to evaluate the obtained results. The final purpose of this study is to build in the near future a large and representative database of human iris spectral reflectances, which can be easily estimated from the images captured by the developed system. This database could be useful in several fields such as medical imaging and the cosmetics industry.