Local weak form geometric active contours for medical image segmentation

We present a local weak form geometric active contour segmentation framework, which naturally unifies the multi-scale parametric and geometric deformable models. This approach makes use of a local weak form formulation of the level set methods, hence inherently allows topological changes during curve evolution. Further, by adaptively selecting the local integration domain for each point of interests, it achieves the strengths of the parametric models of robust boundary detection from noisy or broken edges. Experiment results on synthetic and real medical images provide insights into the superior ability and performance of this strategy.