I am really sorry that I expressed myself wrongly the system looks as follow:

I have got a laser to fiber incoupling unit with two piezomotors for the x and y axis. The output of my fiber is observed by only one photodiode.

If I scan now with the two piezomotors along a N*N big array I would obtain a "map" of values from my photodiode (one value for each x and y position). This map would look a hill (gaussian kurve in 3D) with side maximas (little peaks) but only one main peak.
I want to find this one main peak now without having to scan the whole range of my piezomotors to obtain such a map (because it would take too long) and without getting caught in side maximas (thats why I cannot use the slope between two points algorithm).

I am really sorry and hpe it is clear now.
The suggestion from Steve sounded good and I already got some material - Thank's.

Klaus