Visual saliency estimation for video

Abstract

The most eye catching regions within an image or video can be captured by exploiting characteristics within the human visual system. In this paper we propose a novel method for modeling the visual saliency information in a video sequence. The proposed method incorporates wavelet decomposition and the modeling of the human visual system to capture spatiotemporal saliency information. A unique approach to capture and combine salient motion data with spatial intensity and orientation contrasts in the sequence, is presented. The proposed method shows a superior performance compared to the state-of-the-art existing methods. The fast algorithm can be simply implemented and is useful for many wavelet based applications such as watermarking, compression and fusion.