The role of direction information in the perception of geometric optic flow components

Abstract

Theoretically, optic flow, an important source of information for the perception of locomotion and three-dimensional structure of the environment, is described in terms of divergence, curl, and shear components. We measured how the detection of the type of flow field depends on directional information. We manipulated the local directions by rotating them through an angle x relative to the original direction (i.e., the direction of motion at that locus in an unaltered flow field). The results of the first experiments showed that divergence, curl, and shear can be detected even if the directional range of the individual motion vectors is as broad as 180 degrees. Subsequent experiments revealed that the detection of the geometric components of the optic flow field is merely based on the integration of a few (10% of vectors) local directions correctly (within 10 degrees of original direction) specifying the type of flow field. Other directions are irrelevant to this process. This is actually what one would expect if the optic flow is analyzed by special purpose mechanisms that detect and process the geometric components on the basis of the integration of motion information. The results indicate that as far as they integrate motion information, detectors for divergence, curl, and shear operate in a similar manner. Implications of the results for modeling such mechanisms are discussed.