"Geometry morphometrics" methods (Bookstein 1991) use configurations of landmarks to define the shape of an anatomical structure. Coordinates of these landmarks are then processed to compute an "average" shape, to quantify the variability or to emphasize the differences between data. In general, the considered landmarks are 3D anatomical points which are located either on the real surface of the fossil or on a digital representation. Then it becomes possible to extrapolate "semi-landmarks" in order to define more precisely the shape. Nevertheless, some researchers have proposed to use 3D feature curves directly (Dean 1993) which gives much more information than sparse points. But it can be difficult to define these curves manually and the result remains user-dependent. So, some researchers in computer science have developed methods to extract such kinds of curves automatically (often called "crest" or "ridge" lines) from a 3D image and they have used them to analyse the shape of a fossil skull (Subsol et al. 2002). They also have shown that crest lines are very close to anatomical lines which are extracted under the supervision of an expert. We present the latest algorithms to compute fully automatically crest lines and we apply them on several anatomical structures (tooth, skull, endocranium) of a database of CT-Scan and microCT-Scan images of primates and hominid fossils. We show how these lines emphasize the bony structures of the skull such as the zygomatic process. We find also that crest lines on endocranium data may help to define the different lobes in a reproducible way. Lastly, we use crest lines to characterize the geometry of the grooves in the Enamel Dentine Junction. Based on these preliminary results, we aim to develop new computerized tools to study the morphological variability of anatomical structures and to compare the fossils found on the Sterkfontein site (Braga et al. 2008). Acknowledgements: This study was supported by the HOPE (Human Origins and Past Environments) International Programme funded by the French Embassy in South Africa and the National Research Foundation (South Africa), by the PEPS ODENT Project () funded by CNRS, by the French Ministry of Foreign Affairs, and by the European Commission Marie Curie Research Training Network (European Virtual Anthropology Network; ), according to the Contract MRTNCT-2005-019564 (FP6). The authors thank Stephany Poe for access to the material under her care. References Cited: Bookstein, F. L. 1991. Morphometric tools for landmark data: geometry and biology. Cambridge University Press. Dean D. 1993. The Middle Pleistocene - H. erectus/H. sapiens Transition: New Evidence from Space Curve Statistics. Ph.D. Dissertation, The City University of New York. Subsol G., Mafart B., Silvestre A., and de Lumley M.A. 2002. 3D Image Processing for the Study of the Evolution of the Shape of the Human Skull: Presentation of the Tools and Preliminary Results. In: Three-Dimensional Imaging in Paleoanthropology and Prehistoric Archaeology, Mafart and H. Delingee with the collaboration of G. Subsol (Eds.). British Archaeological Reports International Series 1049, pp. 37-45. Braga J., Subsol G., Thackeray F., Dasgupta G., Balter V., Dedouit F., and Telmon N. 2008. Evolution of Late Pliocene hominin midfacial morphology. An approach using three-dimensional surface registration. American Journal of Physical Anthropology, Suppl. 46: 72.