Abstract:

To model the attention mechanism of human visual system, a visual attention model by fusing spatiotemporal features was proposed. The model is based on the bottomup computational framework, and it uses three low level visual features, including the luminance contrast, the texture complexity and the motion information of foreground objects. By fusing the spatial and temporal features, the visual saliency map of the scene can be achieved. Experimental results show that the proposed model can analyze the saliency of various regions in the sense effectively. Compared with Walther and You models, the proposed model can extract the focus of attention (FOA) in the scene more accurately.

Key words: computer application, visual attention model, feature fusion, saliency map, focus of attention