Entropy is a very important indicator which describes the heterogeneity of the networks. Two kinds of structure entropies, which are based on degree distribution and relative degree respectively, have some problems to describe the characteristics of network structure. Because they focus only on one of node or edge factors. In this paper, We consider the difference between node and edge to define a new network structure entropy. The theoretical analyses and the simulation experiments on regular network, random network and scale-free network prove that this new network structure entropy is more effective to reflect the characteristics of network structure, especially more reasonable to describe the sparse network and star network.