A Population-Based Inference Framework for Feature-Based Attention in Natural Scenes

Abstract. Vision is a crucial sensor. It provides a very rich collection of information about our environment. However, not everything in a visual scene is relevant for the task at hand. Feature-based attention has been suggested for guiding vision towards the objects of interest in a visual search situation. Computational models of visual attention have implemented different concepts of feature-based attention. We will discuss these approaches and present a solution which is based on population-based inference. We illustrate the proposed mechanism with simulations using real world-scenes.

• Type: Conference
• Year: 2005
• Where: BVAI
• Authors: Fred Henrik Hamker
• Keywords: BVAI 2005, Crucial Sensor, Feature-based Attention, Visual