Study Of Visual Selective Attention Models

Information processing is becoming more and more important nowadays.Meanwhile, computer image information processing has been a focus ofattention in modern information era, because traditional manual imageinformation processing is confronted with many difficulties while the amount ofimages that are the main carriers of information at current becomes more andmore prodigious. The human visual system always can identify and take the gistof the scene in front of them quickly. We notice that the interested content isoften a small part of an image which called the salient region in the task ofimage analysis and understanding. Here the salient region can be regarded as themain object in an image, which is the interesting region that human vision canquickly focus on.We present an improved model based on Itti’s saliency model. In this model,different features are filtered with different scales. A model based on top-downmechanisms is presented. We discussed how frequency and color information doin this process.Finally, a model combining bottom-up and top-down mechanisms at an earlystage of visual processing, and predicts the image regions likely to be fixated byhuman observers performing natural search tasks in realworld scenes.Comparing with models based on top-down or bottom-up mechanisms, itperforms as the best predictor of participants’ fixations in the search tasks.