The Study Of Selective Attention Model Based On Human Visual System

Visual information is the main information source for human to perceive,understand and know about the world. The study of visual information processingmechanism is an important part of the brain and cognitive science. In recent years, withthe growing of understanding visual information processing mechanism, more andmore researchers begin to focus on the method of combining visual mechanism withintelligent image processing. Among them, the model of selective attention whichbased on visual mechanism is a hot area. Salient area detection can provide strongsupport for target detection, image compression, helping understanding the calculationprinciple of visual system and so on.Based on information processing mechanism of visual system and the basicmethods of image analysis, the bottom-up visual attention model is deeply studied andanalyzed. Former part of this paper mainly discusses various visual features anddemonstrates that different visual features have different contribution to image scene.Studies have found that people are always attracted by specific areas of an image.These areas mainly consist of some salient features. The method of combiningexperiment and calculation gives us a conclusion that different feature have differentfunction to an image, and provides an effective evaluation method for the followingresearch.Through the study on various image features and traditional attention model of Itti,an improved attention model is brought forward. The main work include: introducinghigh-order image feature (IOC), to reflect the high-order statistics; improving imagepyramid to improve the efficiency of algorithm; using biologically motivated MAXtheory to make better mechanism on features integration. At last our model iscompared with other classical models and proved to be superior.