In an experiment in which two or more photoevents arriving at the same
pixel are considered a single count, a one-pixel-size central hole ap
pears in the autocorrelation estimate. This hole is a source of error
when high-resolution techniques based on the autocorrelation function
are applied to image or signal recovery. We present an experimental me
thod to evaluate the autocorrelation central hole based on the nonline
ar behavior of the total number of counts per frame when clipped photo
count distributions are considered. The technique is easy to perform a
nd provides high accuracy. (C) 1995 Optical Society of America