Computational Visual Attention and Saliency Detection

Visual attention is a mechanism in human perception which selects potentially relevant parts of a scene and focuses processing on these regions. This enables humans to act efficently in their visually highly complex environment. Many computational vision systems face similar problems as humans: a large amount of information has to be processed, usually within a limited time. If the vision system is part of an autonomous system such as a robot, decisions have to be made additionally to determine what to do next: where to look, where to go, what to grasp, etc. Computational attention systems are a biologically-inspired approach to enable a system to decide autonomously which parts of the data are currently of most potential interest within a scene and where to focus processing on.
 
In this talk, I will introduce the state of the art of computational visual attention systems as well as our latest research in this field. Especially, I will focus on saliency detection which has been very popular in the computer vision community during the last decade, and I will introduce our contributions on salient object detection in images, videos, and RGB-D data.