Load balancing algorithms and task scheduling are one of the most important tasks in parallel application design and implementation. Proper task assignment to processor cores can minimize execution time and increase the performance of a parallel application. In this paper, we propose a combined load balancing algorithm based on a mixture of well-known domain decomposition and master-slave algorithms. The proposed algorithm minimizes load imbalance and communication between independent tasks. The proposed algorithm improved parallel efficiency using task rescheduling, which had been confirmed with simulation results.