Shape stylized face caricatures

摘要: Facial caricatures exaggerate key features to emphasize unique structural and personality traits. It is quite a challenge retain the identity of original person despite exaggerations. We find that primitive shapes are well known for representing certain traits, in art psychology literature. Unfortunately, current automated caricature generation techniques ignore role stylization. These methods limited emphasizing distances from fixed Golden Ratio, or computing best mapping proprietary example set (real-image, cartoon portrait) pairs. propose novel stylization algorithm allows expressive vector control with shapes. three shape-inspired ideas input frontal face portraits: 1) Extrapolation Ratio Primitive Shape Spaces; 2) Use stereotype rules; 3) Constrained adaptation supplied mask. adopt recent mesh-less parametric image warp hair, facial (eyes, mouth, eyebrows, nose, ears) provides fast results. The user can synthesize range by changing number constraints, relaxing shape change controlling global exaggeration scaling factor. Different templates rules make person's mimic different personalities, yet basic identity. proposed method easy use implement, be extended create animated games, film interactive media applications.