The Research Of Visual Attention And Quantifying The Relationship Between Visual Saliency And Visual Importance

Visual attention is one of the most important mechanisms in human visual system. In people's everyday life, the environment presents far more perceptual information than the brain can effectively process. In order to keep the essential visual information, human has developed a particular strategy: visual attention. This strategy involves two mechanisms. The first one, which refers to the sensory attention driven by environmental events and low level visual features, is commonly called bottom-up or stimulus-driven. The second one, the attention driven by the task, is usually called top-down or goal-driven. The research of visual attention is based on multiple subjects including computer vision, image processing, biology and psychology. Nowadays, the research of visual attention concerns which part of an image attracts most of people's attention. Peoples develop computational models to get these salient areas.The research of this paper is also the start of my PhD study, some major and creative works includes: (1) introduce the human visual system which provides the biology plausibility of the computational models; (2) introduce of the attention mechanism and talk about bottom-up and top-down mechanism separately; (3) introduce the research progress of computational models and classify the models into 3 classes; (4) introduce two psycho-physics tests which lead to visual saliency (corresponding to bottom-up mechanism) and visual importance (corresponding to top-down mechanism) respectively; (5) quantifying the relationship between visual saliency and visual importance. It is found that the relationship varies according to time and reaches the peak at about 2000 ms. This suggests that a possible HVS strategy of early visual coding might be to quickly locate the main subject(s) in the scene.