A hybrid shape constraint for level set-based active contours

In this paper, we present a new method to incorporate a hybrid shape prior into a level set-based active contour in order to improve its robustness to partial occlusion, clutter and noise. The proposed shape constraint is constructed in two steps. The first one consists in the alignment of the given training data according to the target shape. Then, the Hausdorff distance is used as criteria to choose the most suitable ones to construct the statistical map for being used as a reference model. This constraint is then incorporated, using an evolution schema, into an active contour model to promote the convergence towards the statistical model. Experiments on synthetic and real data show satisfactory segmentation results.