Automatic GrabCut color Image Segmentation Based on EM Algorithm

This paper presents a color image segmentation method using modified GrabCut. GrabCut is considered as one of semi-automatic image segmentation techniques, since it needs user interaction to initialize segmentation process. To eliminate this inconvenience, this paper propose using Expectation Maximization Algorithm for Gaussian mixture model to replace the original initialization process and achieve automatic GrabCut. After experiment and analysis, this paper gives an evaluation over speed and accuracy, proves practicability of this modification and provide a set up for further improvement.