COMPUTER ANIMATION AND VIRTUAL WORLDS Comp. Anim. Virtual Worlds 2007; 18: 279–288 Published online 13 August 2007 in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/cav.200 ........................................................................................... Caricature video By Eun-Jung Lee, Ji-yong Kwon and In-Kwon Lee * .......................................................................... We make moving caricatures from videos on human faces. Using training images, we created a 3D model of an average face. This allows us to transform the image in each frame of an input video, so that it is seen from the front. Then we apply 2D exaggeration rules to caricature each face. Finally, we rotate the face in each frame back to its original position. A panel of viewers gave positive scores to a series of test videos. Copyright © 2007 John Wiley & Sons, Ltd. Received: 18 May 2007; Accepted: 18 May 2007 KEY WORDS: caricature; facial animation; video cartooning Introduction A caricature has been defined as an exaggerated likeness of a person made by emphasizing all of the features that make the person different from everyone else. 1 A caricature makes a more powerful impression than a portrait because of the way that it stresses individuality. Psychological experiment has shown that a caricature can be recognized faster than an accurate picture of a face. 2 Caricatures are widely used in artistic works for fun and in many products. A caricature fulfills not only these aesthetic and commercial purposes but also a psychological purpose. It is not easy for person who does not have the necessary ability to draw a caricature. For that reason, there has been some research on the generation of caricatures using computer graphics. 3–7 These methods potentially allow anyone to generate a caricature. In previous works, caricatures have been made from a single image. We now introduce a technique for generating a moving caricature from an input video which mainly consists of a frontal view of a person’s face. This method can be used to make a video more interesting, as well as for non-photorealistic facial animation, which is often used in constructing games or making animated movies. There are several problems in applying image-based caricature generation methods to video. First, a method that can exaggerate oblique views of the face is needed. Most previous studies of making caricatures have concentrated on the front view, and are not necessarily able to exaggerate other views. We might *Correspondence to: I.-K. Lee, Department of Computer Science, Yonsei University, Seoul, Korea. E-mail: iklee@yonsei.ac.kr use a morphable face model 8 so that the necessary 3D information is available to allow a caricature to be rotated, but this requires 3D scan data from many faces, which are expensive. To solve this problem, we generate a 3D reference model from feature points on training images. Using the resulting model, we can approximate the transformation of the input face. This deformation allows a face can be rotated so that it is seen in an approximately frontal view, and then we can apply existing rules for exaggeration of facial features. Second, we have to create a technique that preserves temporal coherence and is largely unaffected by video noise. The characteristics of the exaggeration method or noise in the input video can cause caricatures to make discontinuous movements between frames when we apply image-based exaggeration methods to each frame of a video separately. In order to protect temporal coherence between frames, we extract an exaggeration parameter from the first frame and then define the exaggeration rules for the remaining frames as simple, continuous linear transformations. The rest of this paper is organized as follows. First, we discuss some of the previous work. Then we present our system framework. After explaining the methods for building a 3D reference model, we present the exaggeration rules next. Then how to generate a caricature video is described. We show some results in the next section and conclude our work and suggest future research in the last section. Related Work Before discussing our method of generating a caricature video, we will review work on the generation of ............................................................................................ Copyright © 2007 John Wiley & Sons, Ltd.