MR Images Segmentation Based on Coupled Geometrical Active Contour Model to Anisotropic Diffusion Filtering

In this paper, a new geometrical active contour model for image segmentation is presented. This model is based on the choice of optimal parameters, in particular in the iterative construction of edge-stopping map function obtained by homogenizing the gray-level of the initial image. To obtain this homogenisation, we propose to couple adaptively the partial derivative equation (PDE) of the anisotropic diffusion process with that of geometrical active contour model. The suggested coupling avoids the leakage problem and ensures that the evolving curve of geometrical active contour model reaches more rapidly the true edge boundary of the objects to be segmented. The performances, in term of robustness and precision of proposed model are evaluated on MR images.