Multiscale Graph-Cut for 3D Segmentation of Compact Objects

The article is a step forward towards improving image segmentation using a popular method called Graph-Cut. We focus on optimizing the algorithm for processing data, in which the target object occupies only a small portion of the total volume. We propose a two-step procedure. At the first step, the location of the object is determined roughly. At the second step, Graph-Cut segmentation is performed with a special multi-scale chart structure. Two different graph construction methods are suggested. The calculation time of both variants is compared with the original Graph-Cut method. The msgc_lo2hi method has been shown to provide a statistically significant time reduction of the computational costs.