Active Contour with Adaptive External Force and Confidential Prior

摘要

Large intra-class deformation, clutter and initialization sensitivity are typical obstacles in image segmentation. To solve these problems, we propose an active contour model with adaptive external force and confidential shape prior. Adaptive external force pushes the evolving curve toward object boundaries independent to initial condition. Confidential shape prior guides the curve by nonuniform strength proportional to the extent of mismatching between prior shape and boundaries. This enables the model to deal with large deformations and clutter. Experiments and comparison with other deformable models show advantages of our model.