| Visual search is an important strategy for human visual system to realize fast scene perception.Guided search theory holds that global layout or other top-down scene sources play an important role in guiding target search.In order to verify the specific role of scene layout and area cues in guiding visual attention,this paper carried out the following two aspects of research.In the first study,we used eye tracking system to do psychophysical experiments on the outline of natural scenes.We collected the fixation images of 10 subjects in498 natural images and 996 boundary contours of human markers(two boundary images for each image).The experimental results show that in the absence of color,brightness and other basic features,the distribution of fixation points on the contour line is highly correlated with that on the natural image.In addition,compared with the basic cues of the area,the subjects pay more attention to the closed areas of the contour,which are usually related to the main objects of the scene.The second research content is to establish a computational model to predict the fixation area.The experimental results of study 1 show that the closed area is more likely to form meaningful objects and attract most of our visual attention.Therefore,we build a simple model to extract the closed area and generate a priori mark where the object may appear.This priori will be used to guide the generation of bottom-up significance and improve the accuracy of fixation prediction.This model proves that in the natural scene,the fixation information on the contour line can significantly improve the fixation prediction performance of the classic bottom-up model.These results support that Gestalt features and scene layout are important guiding clues for fast visual target search.The experimental results show that using contour information can significantly improve the performance of the classic bottom-up method for natural scene gaze prediction. |
PDF Full Text Request