Selective visual attention system based on spatiotemporal features

Abstract

In this paper, a selective visual attention system for motion pictures is proposed. The proposed attention system is new in that it utilizes motion information for the purpose of detecting Region Of Interest (ROI) or Focus Of Attention (FOA) in motion pictures. Typical feature integration model is expanded to incorporate motion stimulus in our suggested model. Suggested model is able to respond to motion stimulus by employing motion fields as one of temporal features to the feature integration model. Analysis of motion field maps and incorporation of the result are distinct from some of the previous studies on spatial feature integration. Comparative experiments with a human subjective evaluation show that correct detection rate of visual attention regions improves by utilizing temporal features compared to the case of using only spatial features.