Iterated graph cuts with confident measure
Dongliang Yang, Tingquan Deng · 2011
In this paper, an iterated graph cuts based image segmentation approach is proposed. Graph cuts method [1] obtains segmentation in an iterative version of optimization framework. However, the graph cuts algorithm may not segment object well because of much interference from inaccurate updated models. The proposed method works with the new updated models of object to reduce the interference significantly. A novel strategy is proposed to update object models, thereby high confident components can be selected using a new confident measure (CM). The experimental performance demonstrates the validity and effectiveness of the proposed method.