How does the Hessian transformation work?

edit

I am trying to wrap my head around OpenCV's SURF algorithm. For the algorithm, points of interest are calculated by creating a Hessian Matrix around each pixel. But what does this mean? I understand the hessian matrix is just taking the 2nd derivatives - but why does it work? What is the significance?

I know this is more a math question than an OpenCV one, but this stems from an OpenCV project. Thanks for your answers, J