In this paper we propose a simple new algorithm to perform clustering, based on the Alter proposed in [1] but lowering significantly algorithmic complexity with respect number of clusters. An empirical study states relevance our iterative process and confrontation simulated multivariate functional data shows benefits algorithm.