The role of natural image statistics in biological motion estimation

Abstract

While a great deal of experimental evidence supports the Reichardt correlator as a mechanism for biological motion detection, the correlator does not signal true image velocity. This study examines the accuracy with which physiological Reichardt correlators can provide velocity estimates in an organism’s natural visual environment. Both simulations and analysis show that the predictable statistics of natural images imply a consistent correspondence between mean correlator response and velocity, allowing the otherwise ambiguous Reichardt correlator to act as a practical velocity estimator. A computer vision system may likewise be able totak e advantage of natural image statistics toac hieve superior performance in real-world settings.