GrabCut color image segmentation based on region of interest
Siyang Hua, Ping An Shi · 2014
In order to speed up the convergence of GrabCut algorithm and enhance the accuracy of segmentation, this paper puts forward an algorithm of GrabCut color image segmentation based on region of interest (ROI). The user selects a ROI by dragging a rectangle. The GrabCut algorithm is used in ROI only. The object is then extracted automatically. The experimental results demonstrate that the new algorithm costs less and is more accurate than the original GrabCut algorithm for images with complex background.