Regularity-Based Caricature Synthesis

In this paper we present a regularity-based caricature synthesis technique from a frontal face image, consisting of shape exaggeration, relationship exaggeration. Three rules are introduced which are important in caricature exaggeration. The first is the relativity concept: the relationship between features of different subjects and the relationship between facial features of the same subject, which make the exaggerative effect more contrastive. The second is the so called proportion, which characterises the face in a proportion form. It makes global feature synthesis a simple task and the normalizing of facial features unnecessary. The third is hierarchical-exaggeration model. We define all facial proportions three hierarchy to guide exaggeration. The effectiveness of our algorithm is demonstrated with experimental results.