Human Systems Integration Division @ NASA Ames

It is commonly assumed that the perception of motion transparency arises from the activity of multiple channels tuned for spatial and temporal frequency, which receive input from the same portion of visual space; this belief is supported by observations with sine-wave plaid patterns where the component frequency is varied. We have investigated this hypothesis using a class of stimuli which are derived from two dimensional noise patterns. The patterns are first filtered for various combinations of spatial frequency and orientation. Composite stimuli are produced by algebraic summation of the generating patterns, which are independently scrolled on a frame-by-frame basis. A basic finding is that it is difficult to evoke the percept of transparency when the stimulus is confined to a narrow band, either of spatial frequency or orientation.

A computer model has been developed in which motion information extracted by a linear sensor is represented by a spatiotopically organized layer of units tuned for spatial frequency, orientation, and direction of motion. To mimic observers responses it is necessary to embellish the model by the addition of additional features:

1) spatial spreading and temporal integration of responses;
2) cooperative interactions between direction selective units tuned for different orientations.