|
Vive la différence ! Intégration et différentiation dans le système visuel| old_uid | 347 |
|---|
| title | Vive la différence ! Intégration et différentiation dans le système visuel |
|---|
| start_date | 2005/12/09 |
|---|
| schedule | 11h |
|---|
| online | no |
|---|
| summary | A critical step in the interpretation of the visual world is the integration of the various local motion signals generated by moving objects. This process is complicated by the fact that local velocity measurements can differ depending on contour orientation and spatial position. Specifically, any local motion detector can measure only the component of motion perpendicular to a contour that extends beyond its field of view. As a result of this "aperture problem," velocity measurements made along the edge of an object are ambiguous, whereas those made at corners, or other regions of contour discontinuity, are not. This problem is particularly relevant to direction-selective neurons early in the visual pathways, where small receptive fields permit only a limited view of a moving object.
I will describe experiments showing that neurons in the middle temporal visual area (MT/V5) of alert monkeys reveal a dynamic solution to the aperture problem. MT neurons initially respond primarily to the component of motion perpendicular to a contour's orientation, but the responses gradually shift, over a period of ~100 ms, to encode the true stimulus direction, regardless of orientation. One possible mechanism for this solution involves selectivity for the motion of contour discontinuities, or "terminators," which are known to strongly influence motion perception. To examine this possibility further, we measured neuronal responses to "barber-pole" stimuli, in which the aspect ratio of a grating aperture is used to vary the relative proportions of terminators moving in orthogonal directions. Consistent with perceptual effects, we found that the sustained directional responses of MT neurons are very close to those predicted by a vector-average of terminator signals only. This emphasis on terminator motion may result from an earlier stage of visual processing in which V1 neurons are rendered less sensitive to contours through end-stopping. We tested this possibility by characterizing V1 responses in alert macaques using sparse, "white noise" stimuli composed of long bars at various orientations. We used reverse correlation to measure both the spatial (1-bar) and directional (2-bar) interactions from the same spike trains. The results showed that end-stopped, direction- selective neurons in V1 provide reliable 2-dimensional motion signals that are relatively unaffected by bar orientation. In addition, a time-slice analysis of the interactions revealed that end-stopping itself takes some time to evolve: at short latencies (~40 ms after stimulus onset), the neurons responded well to a long bar regardless of it position within their receptive fields, whereas at longer latencies (~80 ms) they responded only to the bar's end-points. This suggests that at least some of the dynamics of the solution of the aperture problem seen in MT neurons may be accounted for by the dynamics of end-stopping in its inputs. |
|---|
| responsibles | Riehle |
|---|
| |
|