System decomposition has become a research focus in software maintenance and evolution for a long time. In this paper, we propose a hierarchical decomposition method for object oriented systems based on identification of omnipresent clusters. The distinctive features of this approach are as follow: Firstly, we addressed the problem of omnipresent classes for class diagram. Secondly, we proposed a hierarchical decomposition strategy which can deal with unbalanced hierarchical organization for software system. Thirdly, we presented a revised independency metric that can offset the impact of the measured clusters? size. The experimental results show that this method can be both effective and efficient.