Application of stochastic counterpart optimization to contrast-detection autofocusing

A simple model for the contrast-detection autofocusing problem is considered. The variance of the image is examined as a focus function. We prove that the standard convergence rate of the variance estimate (empirical focus function) of order O(T
^-1
), where T is a (relative) sensor size, allows direct application of the golden-section search algorithm to the empirical focus function.