Information networks are ubiquitous. These networks can be modeled by graphs where nodes refer to objects and edges are relationships between objects in the networks. Measuring nodes similarity in a graph is a fundamental problem of graph data management. There are many real-world applications based on similarity between objects, such as networks analyses, information retrieval and recommendation systems. A number of link-based similarity measures have been proposed. Both SimRank and personalized PageRank are the most popular and influential similarity measures. The two measures differ from each other because they depend on different paths. This paper proposes a similarity measure that takes advantages of both SimRank and personalized PageRank. The new measure strengthens the principle of SimRank: "Two objects are similar if they are referenced by similar objects". The paper also analyzes the new similarity measure in theory and proposes an optimization algorithm which reduces the time cost from O(kn⁴) to O(knl) where k is the number of iteration, n is the number of nodes and l is the number of edges. Experimental results demonstrate the effectiveness of the new similarity measure and efficiency of the optimization algorithm.