In this article we propose an original method for the anatomical labeling of vertebrae from 3D CT-scan images. The primary purpose of this work is to obtain a robust referential of the abdomen. This referential can be used to locate anatomical structures like organs or blood vessels. The main problematic concerns the separation of the vertebrae, which are structures that are very close from each other. In order to detect the intervertebral spaces, we use a morphological operator which detects the dark spaces corresponding to intervertebral discs in combination with an analysis of the shape of the vertebrae in the axial plane. To reconstruct the vertebrae we use the paradigm of mathematical morphology, which consists in finding markers inside the vertebrae and compute the watershed from markers. Then we label the vertebrae according to their anatomical names. To do this, we automatically detect T12 vertebrae. We have evaluated our algorithm on 26 images.