Image Segmentation Based On Salient Object Detection And Its Applied Research

Image segmentation has a wide range of applications in image processing,and it has been involved in various image types.Moreover,salient object detection can provide reliable and valuable information for image segmentation,so it’s of great importance to effectively combine salient object detection with image segmentation.In this paper we study the problems of image segmentation based on salient object detection and apply this method to the segmentation of magnetic resonance images with intensity inhomogeneities.GrabCut algorithm can successfully associate saliency detection with image segmentation.Firstly we receive the saliency map by salient object detection algorithm,secondly we can use proper threshold value to get target object contour,lastly take the object contour as the input of GrabCut algorithm,so accurate segmentation result can be obtained.In this paper,we separately take advantages of Kapur Entropy and Maximum Entropy to get the object contour,then set the contour as GrabCut’s input,and we conclude that both methods have pros and cons,and the Maximum Entropy method combined with GrabCut gets more complete segmentation outline but fails in local details.In addition,we apply the image segmentation method combined with salient object detection to medical images segmentation in the presence of intensity inhomogeneities.What’s more,the method above can be separately solved by the variational method and level set method.Finally experimental comparison on both synthetic images and real MRI data show that this method achieves better in terms of accuracy and robustness.