Implementation of visual attention system using bottom-up saliency map model

Abstract

We propose a new active vision system that mimics human-like bottom-up visual attention using saliency map model based on independent component analysis. We consider the feature bases reflecting the biological features and psychological effect to construct the saliency map model, and the independent component analysis is used for integration of the feature bases to implement human-like visual attention system. Using the CCD camera, a DSP board, and DC motors with PID controllers, we implement an active vision system that can automatically select a visual attention area. © Springer-Verlag Berlin Heidelberg 2003.