A Note on Local Min-Max affine approximations of real-valued convex functions from R^k with applications to computer vision

We present a method to find a local Min-Max affine approximant to a given convex function $f:R^k-R$, and show an application to computer vision/graphics.