Rendering and animating expressive caricatures
Mohammad Obaid, Ramakrishnan Mukundan, Mark Billinghurst · 2010
We present a novel approach for generating stylized artistic rendering of caricatures from a given face image, with the ability to map any one of the main six expressions and control the degree of its expressiveness on the generated caricature. Our method achieves this by manipulating the facial appearance and expressivity of the caricature, using quadratic deformation model representations of facial expressions. A stroke-based non-photorealistic rendering (NPR) engine is developed to generate the output caricature that appears to be a sketch of the original image. The NPR engine segments the defined rendering path and generates rectangular strokes that approximate each segment. The stroke properties are computed by applying geometric image moments on each segment. To enhance the overall appearance of the caricature, extra strokes are randomly positioned around the rendering path. The results show that from an input neutral facial image, expressive rendered caricatures are generated for the facial expressions neutral, smile, sad, fear, surprise, disgust, and anger. Our approach produces very expressive artistically rendered caricatures, and could lead to future research directions in further extending the method to produce pencil sketch rendering of images or rendering 3D characters to generate cartoon-like 3D characters.