The Study Of Salient Object Detection Algorithm Based On The Visual Perception Mechanism

With the development and progress of the times,image and video has become an indispensable and important media and means for us to know the outside information.So,how to deal with these information properly have become particularly important.Vision as the most important means for our human to have access to the external information,is superior to any existing computer vision system in all aspects of information processing.Therefore,it has aroused widespread concern that how to apply visual system perception and processing of external information mechanism to the actual processing of the computer.For a long time,studies in the fields of psychology,neuroscience,and computer vision have shown that the human brain and the visual nervous system have more attention to some objects in the image,and we call these goals as salient objects.Salient object detection is not only an important task for human visual perception,but also for computer vision applications.Salient objects are widely used in engineering applications such as target detection and identification,image and video compression,visual tracking,and human-computer interaction and so on.In these applications,we can better retain the object information by using a salient object.In this paper,we mainly establish salient object detection methods,which is based on the visual perception mechanism from the perspective of the model,for the study of image's targets detection algorithm.The content of the paper is divided into the following two parts:(1)Contrary to the popular salient object detection methods which are based on pixel or super-pixel,we propose a salient object detection method based on segmentation,and use the existing high quality segmentation method to simplify the salient object detection.For the input image,firstly,we use existing segmentation method(g Pb-OWT-UCM)to segment the image;And then we compute the three very simple image features of the segmented image.We can easily extract the salient object from the segmented image.Finally,we compare our method to some of the existing algorithms in some databases(e.g.ASD),and our method has a good performance.(2)Another research in this paper is to propose a salient object detection method based on the Gestalt theory.We analyze and understand the process of the human visual system to process information,we make some improvement of the existing model based on visual search guidance theory in visual perception mechanism and the Gestalt theory.In the non-selective path,we use the Gestalt principles to get the target area's position by computing some image structure features(in the global),which is used as a prior information;In the selective path,we extract the local features of the input image,e.g.brightness,color,texture.Finally,we use the Bayesian theory to integrate two kinds of information and use automatic iteration to terminate the operation to get the final saliency map.Then,we compare our method to some of the existing algorithms in some databases(e.g.ASD),analyze and summarize its advantages and disadvantages.Finally,we did two validation experiments to analyze the performance and feasibility of our method.