This paper presents an efficient method to reduce the optimization cost. In this method, the eigenvectors of the Hessian of the objective function are determined first. Then, the search for the optimum is carried out successively in subspaces defined by these vectors. For this purpose, the Multi-directional-Search Algorithm is used in this study, but any other optimization algorithm can be employed. The method is validated in two test cases: analytical function and shape reconstruction problem. In both cases, this method shows very promising results.