John Tsotsos
Department of Computer Science and Center for Vision Research, York University

Can Distributed Local Saliency Computations Solve the Feature Binding Problem?

Friday 04th of November 2005 at 02:30pm
5101 Tolman

Computational vision has a long history of proposing methods for decomposing a visual signal into components. For example, many good strategies have appeared for decomposing visual motion signals (such as Heeger, Sperling). What has been far more elusive is how to recombine those components into a whole. This problem has even merited its own name - the binding problem. To date no realizable process has appeared to solve the binding problem, even in part, although several proposals have appeared. This paper proposes a novel solution for a significant portion of the binding problem, namely, the re-combination of visual features into larger patterns and their localization in the image. The solution requires the abandonment of the nearly ubiquitous saliency map and adoption of a distributed, localized computation of saliency that is dependent on local neural selectivity constraints. This strategy is demonstrated within a fully implemented model that attends to simple motion patterns in image sequences.

Join Email List

You can subscribe to our weekly seminar email list by sending an email to majordomo@lists.berkeley.edu that contains the words subscribe redwood in the body of the message.
(Note: The subject line can be arbitrary and will be ignored)