Detecting moving objects in an optic flow field using direction- and speed-tuned operators

- We present a model for detection of moving objects in an optic flow field.
- The model uses operators that are tuned to direction and speed of motion.
- The model accurately detects moving objects without explicitly computing speed.

An observer moving through a scene must be able to identify moving objects. Psychophysical results have shown that people can identify moving objects based on the speed or direction of their movement relative to the optic flow field generated by the observer's motion. Here we show that a model that uses speed- and direction-tuned units, whose responses are based on the response properties of cells in the primate visual cortex, can successfully identify the borders of moving objects in a scene through which an observer is moving.