Multiphoton microscopy (MPM) is a tool highly involved in biological imaging. Its physical principle is associated to a reduced resolution compared to more standard systems: a micrometer in radial direction and few micrometers in axial direction. Here, we propose a computational approach in order to restore multiphoton images. The pipeline processing strategy is gathered into our algorithme FIGARO where microbeads are imaged and estimated. The tartegy is besed on a variational approach in order to generate a multivariate gaussian fitting procedure presenting a high level of robustness to noise and blur, indispensable in MPM.