For separating linear instantaneous mixtures of independent/dependent source components, we extend the independent/dependent blind source separation method, proposed by Keziou et al. (2014), to cover the more general case, where the dependency structure of the source components and the related parameter are both unknown. An application is given for separating scanned recto-verso documents.

Blind source separation; Copulas; Kullback-Leibler divergence between copula densities