Multi-level 3D caricatures with relative feature displacement and head shapes

Facial caricatures have a many uses ranging from entertainment to psychological research but they are difficult and costly to produce in terms of both time and money. Many 2D automatic caricature generators have been presented but these are invariably limited to frontal portraits whereas traditional caricaturists usually choose a three-quarter view to best convey form. A 3D system would allow unlimited viewpoints but current 3D systems have had limited success. In this paper we present a system that automatically generates expressive 3D caricatures from orthographic input images. By analysing the methods used by traditional caricaturists we have identified techniques that can be easily applied to 3D caricatures. By blending techniques for head modelling, 2D caricature generation, and traditional caricature [17, 7, 19] we have implemented a system that allows users to quickly create expressive 3D head models. We describe a multi-level system that incorporates the traditional caricaturing concepts of Relative Feature Displacement and Head Shapes so that the artefacts produced more closely mimic the techniques used by caricature artists