|Biological Plausibility of Discriminant Tracking|
We show that saliency based discriminant tracking, is biologically plausible, by mapping the tracker into a network compliant with the widely accepted neurophysiological models of areas V1 and MT. This mapping extends the substantial connections between discriminant saliency and the standard model that have already been shown.
In particular, we show that all information required for optimal (in a decision-theoretic sense) feature selection can be obtained by divisive normalization, across feature channels, of the responses of the saliency network. The resulting network is a biologically plausible optimal model for both saliency and tracking.
Further, we show that the tuning of top-down saliency associated with this feature selection mechanism explains the well-known
Finally we present results of several psychophysics experiments on the dependence between target saliency and human tracking performance, and demonstrate that the saliency based tracker is compliant with this data. These experiments build on well understood properties of saliency, such as pop-out effects, to show that tracking requires discrimination between target and background. In addition, we characterize the dependence of tracking performance on the extent of discrimination by gradually varying feature contrast between target and distractors in the tracking tasks. The results show that both tracking performance and saliency show highly similar patterns of dependency on feature contrast. This provides strong evidence for the proposed connection between saliency and tracking.
V. Mahadevan,and N. Vasconcelos.
In preparation. Journal of Vision,
|Contact:||Vijay Mahadevan, Nuno Vasconcelos|