Topology-preserving 3D image segmentation based on hyperelastic regularization (Q2028532)

Image segmentation is partitioning a digital image into multiple segments for extracting meaningful objects from the image. Noise (e.g. in medical images) is reduced using a prior information, here the topological structure of the target object (registration). Here, the proposed by \textit{H.-L. Chan} et al. [J. Math. Imaging Vis. 60, No. 3, 401--421 (2018; Zbl 1437.94011)] topology-preserving image segmentation using the Beltrami representation of shapes (suitable only for 2D images) is modified replacing the Beltrami regularizer with hyperelastic regularization proposed by \textit{M. Burger} et al. [SIAM J. Sci. Comput. 35, No. 1, B132--B148 (2013; Zbl 1318.92028)], which works also with 3D images. It is shown (theoretically), that the proposed model is solvable and a converging generalized Gauss-Newton iterative scheme is proposed to solve the model. The effectiveness of proposed method is illustrated by numerical experiments with synthetic and real images.