In this paper we propose a multi-resolution visualization approach based on a forcedirected algorithm to graphically and hierarchically represent large-scale data sets. Considering the limitations of the human visual cognitive system, it is reasonable to dynamically and interactively extract the most important or user-specified information for representation. To smoothly animate the merging or splitting of nodes and edges in a graph, we adopt a jelly-like metaball technique. Our prototype system clusters and graphically represents thousands of nodes and links in a novel way that allows a user to recognize visually. By adopting our system, the hierarchy of information and the evolving trends are dynamically visualized through an interactive 3D graphical interface, in addition to simultaneous display of data detail and summarization. Furthermore, user perception can be enhanced by adopting human 3D stereopsis skills such as Perspective, Chiaroscuro, and Focusing method