On the recognition of order and disorder.

We compare several algorithms for the recognition of ordered and disordered images. As image sources we use waves of the Belousov-Zhabotinskii reaction coupled to convective motion in a petri dish. This device allows reversibly the generation of periodic (ordered) and aperiodic (disordered) patterns. The best match between the parametric description and the observations is obtained by an "autodifference function". This function is computed by summing up intensity differences over all pairs of picture elements having a given distance on the picture plane. Then, the minimum of this function is determined upon variation of the distance. This algorithm is not only efficient for the recognition of order and disorder in "machine vision", but also plausible in biological visual perception.