Multivariate methods have been recently introduced and successfully applied for the discrimination of signal from background in the selection of genuine very-high energy gamma-ray events with the H.E.S.S. Imaging Atmospheric Cerenkov Telescope. The complementary performance of three independent reconstruction methods developed for the H.E.S.S. data analysis, namely Hillas, model and 3D-model suggests the optimization of their combination through the application of a resulting efficient multivariate estimator. In this work the boosted decision tree method is proposed leading to a significant increase in the signal over background ratio compared to the standard approaches. The improved sensitivity is also demonstrated through a comparative analysis of a set of benchmark astrophysical sources.