ABSTRACT
Whenever an animal is moving in its environment, moves its eyes or an object moves in front of its eyes, the visual system is confronted with motion. However, this motion information is not explicitly represented in the two-dimensional brightness pattern of the retinal image. Instead, motion has to be computed from the temporal brightness changes in the retinal image. This is one of the first and most basic processing steps performed by the visual system. This primary process of motion detection has become a key issue in computational neuroscience, because it represents a neural computation well described at the algorithmic level that has not been under-stood at the cellular level in any species so far, yet simple enough to be optimistic in this respect for the future. The development of models of motion detection has been experimentally driven in particular by investigations on two systems, the rabbit retina [1] and the insect visual system [49]. Vice versa, there is probably no other field in system neuroscience where experiments were more influenced by theory than in the 1-58488-362-6/04/$0.00+$ 1.50
study of motion vision. Motion vision has thus become a classical problem in computational neuroscience, which many laboratories around the world have embarked on. In the following I will give an overview of what is known about the computations underlying motion vision in the fly, where a lot of experimental results are available (for review see: [13, 36]) and where modelling efforts have reached a rather detailed biophysical level at many processing steps.
