Congestion often occurs due to the data overload in real network. Central nodes of the network are firstly congested and then will lead to the entire network congestion. Central nodes become the bottleneck of network capacity. In order to balance the traffic load of central nodes and improve network capacity, we propose a link removing strategy that removing a fraction of links in the network according to both the degree and the betweenness of the nodes. Extensive simulations under the shortest path routing are carried out for the effectiveness of the proposed strategy. It is found that our proposed strategy can bring higher traffic capacity than the high-degree-first (HDF) strategy and high-betweenness-first (HBF) strategy. What we have done is helpful for network design and network performance optimization.