Abstract | Keywords | Export | Availability | Bookmark

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

The face is one of the most characteristic features in humans. In the fields of forensics and anthropology it can be interesting to infer the face from the underlying bone structure, as most of the times only the skull remains. When the bone is deformed or incomplete, virtual skull reconstruction can be a useful tool. Starting with a dataset of skull meshes, a model is built. The model is adapted to the shape of the skull with defects giving a new one that resembles the original, but without the defects. In the current work two datasets of skulls from different origins were available: African and West European. The African dataset was of specific interest because the samples already contained defects. Three objectives were established and accomplished: define, implement and validate an algorithm for virtual skull reconstruction; build a model based on data with defects and establish the importance of race for skull models. Despite unwanted variation appeared in regions where defects were frequent in the original dataset (especially the teeth), a qualitative analysis lead to the conclusion that the model was satisfactory. The importance of race was stated because unavoidable artifacts were generated in the reconstruction of African skulls when using a model based on the West European dataset. A quantitative analysis was proposed, measuring the distances between correspondent points of the reconstructed and original skull meshes. The quantitative results showed better performance when a second step of non-rigid registration was applied after model adaptation. It was concluded that samples without defects are needed for accurate quantitative validation, even for validation of a model based on data with defects.