The topic of this thesis is automatic analysis of craniofacial images. The methods proposed and applied contribute to the scientific knowledge about different craniofacial anomalies, in addition to providing tools for detailed and robust analysis of craniofacial images for clinical and research purposes. The basis for most of the applications is non-rigid image registration. This approach brings one image into the coordinate system of another resulting in a deformation field describing the anatomical correspondence between the two images. A computational atlas representing the average anatomy of a group may be constructed and brought into correspondence with a set of images of interest. Having established such a correspondence, various analyses may be carried out. This thesis discusses two types of such analyses, i.e. statistical deformation models and novel approaches for the quantification of asymmetry. The analyses are applied to the study of three different craniofacial anomalies. The craniofacial applications include studies of Crouzon syndrome (in mice), unicoronal synostosis plagiocephaly and deformational plagiocephaly. Using the proposed methods, the thesis reveals novel findings about the craniofacial morphology and asymmetry of Crouzon mice. Moreover, a method to plan and evaluate treatment of children with deformational plagiocephaly, based on asymmetry assessment, is established. Finally, asymmetry in children with unicoronal synostosis is automatically assessed, confirming previous results based on manual reference points and providing a higher level of detail.