New tensorial morphological gradients in DTI image segmentation algorithm

Abstract

Calculating morphological gradient is the key step of watershed algorithm. In this article, new tensorial similarity morphological gradient is defined based on the eight neighborhoods, and new tensor anisotropy morphological gradients are put forward, which are then used in the watershed segmentation framework to segment DTI image. The results of the segmentation experiments on the corpus callosum show that: Compared to other tensor anisotropy morphological gradients based watershed segmentation methods, the one based on newly proposed tensor anisotropy morphological gradients can more quickly and accurately position and depict the segmentation outline of the image, which can also better protect the edge information of the important region.