Maximum of a noised surface fitting paraboloid

Hey,

I'm working on a surface that presents a global maximum, and my goal is to get the indexes of this maximum. As there is a small noise on the surface, my idea is to locally fit a paraboloid to be more precise. That could also be helpfull in the case where my surface is very flat close to the maximum. That may also be seen as an interpolation.

Does anyone know any other method than fitting this paraboloid to help me with this problem (noise + flat maximum) ?