Many algorithms have been developed to solve the problem of detecting and analyzing community structure in networks. Louvain algorithm (LVA) is a well-known community detection method that results high community structure of large networks within reasonable time, but it has the problems of randomness and instability. In this paper, an improved Louvain algorithm (ILVA) is proposed by combining the modularity function and node importance with the original LVA. The ILVA uses the LVA to detect community structure by optimizing the value of modularity. Meanwhile, node importance as measured by degree centrality is used to determine the node scanning order in the community detection phase. Experiments were conducted on real-world networks and the results showed that the ILVA produced stable community structure with higher modularity within reasonable time.