Level set constrained segmentation using local curvature

A novel method for level set based segmentation of images using constrains depending on image characteristics is presented. Our method is motivated by the fact that due to the unhomogeneity of image regions, segmentation algorithms often fail because the final contour should depend in those regions. We propose to add constrains, to level set equation, depending in image characteristics. We apply a local coefficient depending on curvature in a local neighbourhood at each point. So the final contour is more homogenous with smoothed regions and more curved regions. In addition, as the stop forces in basic level set equation are not enough to stop propagation in weak edges, we propose to begin by adding surface minimization and region intensity constrains to improve the propagation of the final contour. The surface minimization term proves stabilization efficiency even in presence of nosy and weak boundary. Results illustrated with a medical image demonstrate the efficiency of the method.