Dynamic load balancing is an important step conditioning the performance of parallel adaptive codes whose load evolution is difficult to predict. Most of the studies which answer this problem perform well, but are limited to an initially fixed number of processors which is not modified at runtime. These approaches can be very inefficient, especially in terms of resource consumption. In this paper, we present a new graph repartitioning algorithm which accepts to dynamically change the number of processors, assuming the load is already balanced. Our algorithm minimizes both data communication and data migration overheads, while maintaining the computational load balanced. This algorithm is based on a theoretical result, that constructs optimal communication patterns with both a minimum migration volume and a minimum number of communications. An experimental study which compares our work against state-of-the-art approaches is presented.