Abstract: In this paper, inspired by the findings in cognition psychology, we study the data competition in the bottom-up visual attention system and bring up a novel visual attention model based on the hierarchical competition. By this model, the competitions of scale, feature and location are applied hierarchically to define the fundamental attributes of the focus of attention (FOA), and then the method of region growth is employed to describe the contour of the FOA. In this way, every FOA with different properties in an image is detected in series. We apply the model to various real images, and the exciting outcome shows that the model is effective.