Motion information is required for the solution of many complex tasks of the visual system such as depth perception by motion parallax and figure/groundThinking about how the nervous system extracts motion information, one might imagine that it compares successive images to measure the displacement of a certain object. This implies that the identification of specific features in a scene is a prerequisite for the perception of motion. Although feature identification may play a role in motion vision under certain circumstances a, it is not necessary. This could be shown in psychophysical and behavioural experiments on various species by the use of periodic or statistical patterns containing no prominent features a'2. It is now widely accepted on the basis of such experiments that motion is initially evaluated in parallel by two-dimensional, retinotopically organized arrays of local motion detectors, which operate in the simplest case directly on the local light intensity (for review see Ref.3). Models of motion detectionFrom a theoretical point of view, local motion detection mechanisms have to satisfy certain minimum requirements in order to signal motion in a directionally selective way (see Box 1). In brief, a movement detector has to be asymmetrical and needs at least two inputs, which interact in a non-linear way 4-6. These requirements are met by a variety of similar motion detection models 6-1°. Some of them characterize the computations underlying motion detection in formal terms, others try to account for them in terms of cellular mechanisms. Irrespective of the actual level of description, the various biological motion detection schemes have been divided into two main categories, the so-called gradient-and correlation-type models 8'H. While in the gradient schemes an estimate of local motion is obtained by relating the simultaneously measured spatial and temporal changes in local light intensity of the moving image7,S, la, in the correlation schemes 2'6A2 and their mathematical equivalents 13 this is done by evaluating a kind of spatiotemporal cross-correlation of the appropriately filtered signals originating from two points in the retinal image. The gradient scheme originated from the study of computer vision and was applied only later to biological motion vision 7's. In contrast, the correlation-type of movement detector was deduced from behavioural experiments on motion vision in insects 2'6. Subsequently, it was successfully used to explain motion detection in vertebrates including man 12'14-a8. The well-known 'BarlowLevick'-type of movement detector, which was originally proposed to account for motion detection in the rabbit retina 19, belongs also to the broad class of correlation-type movement detectors in that it is a variant of them, being effected by logical operations rather than in an analogue form. Two Inputs Two inputs are necessary since motion is a vector that needs two points for its representation. A single photoreceptor could not distinguish a dark bar that crossed its receptive field...

No abstract

Krapp, Holger G., Roland Hengstenberg, and Martin Egelhaaf. Binocular contributions to optic flow processing in the fly visual system. J Neurophysiol 85: 724 -734, 2001. Integrating binocular motion information tunes wide-field direction-selective neurons in the fly optic lobe to respond preferentially to specific optic flow fields. This is shown by measuring the local preferred directions (LPDs) and local motion sensitivities (LMSs) at many positions within the receptive fields of three types of anatomically identifiable lobula plate tangential neurons: the three horizontal system (HS) neurons, the two centrifugal horizontal (CH) neurons, and three heterolateral connecting elements. The latter impart to two of the HS and to both CH neurons a sensitivity to motion from the contralateral visual field. Thus in two HS neurons and both CH neurons, the response field comprises part of the ipsi-and contralateral visual hemispheres. The distributions of LPDs within the binocular response fields of each neuron show marked similarities to the optic flow fields created by particular types of self-movements of the fly. Based on the characteristic distributions of local preferred directions and motion sensitivities within the response fields, the functional role of the respective neurons in the context of behaviorally relevant processing of visual wide-field motion is discussed.

No abstract

The computations performed by individual movement detectors are analyzed by intracellularly recording from an identified direction-selective motion-sensitive interneuron in the fly's brain and by comparing these results with model predictions based on movement detectors of the correlation type. Three main conclusions were drawn with respect to the movement-detection system of the fly: (1) The essential nonlinear interaction between the two movement-detector input channels can be characterized formally by a mathematically almost perfect multiplication process. (2) Even at high contrasts no significant nonlinearities seem to distort the time course of the movement-detector input signals. (3) The movement detectors of the fly are not perfectly antisymmetrical; i.e., they respond with different time courses and amplitudes to motion in their preferred and null directions. As a consequence of this property, the motion detectors can respond to some degree to stationary patterns whose brightness is modulated in time. Moreover, the direction selectivity, i.e., the relative difference of the responses to motion in the preferred and null directions, depends on the contrast and on the spatial-frequency content of the stimulus pattern.