Arts and Technology. First International Conference, ArtsIT 2009, Yi-Lan, Taiwan, September 24-25, 2009, Revised Selected Papers

Research Article

Automatic Generation of Caricatures with Multiple Expressions Using Transformative Approach

Download
487 downloads
  • @INPROCEEDINGS{10.1007/978-3-642-11577-6_33,
        author={Wen-Hung Liao and Chien-An Lai},
        title={Automatic Generation of Caricatures with Multiple Expressions Using Transformative Approach},
        proceedings={Arts and Technology. First International Conference, ArtsIT 2009, Yi-Lan, Taiwan, September 24-25, 2009, Revised Selected Papers},
        proceedings_a={ARTSIT},
        year={2012},
        month={5},
        keywords={Caricature generation facial animation parameters image morphing facial feature localization},
        doi={10.1007/978-3-642-11577-6_33}
    }
    
  • Wen-Hung Liao
    Chien-An Lai
    Year: 2012
    Automatic Generation of Caricatures with Multiple Expressions Using Transformative Approach
    ARTSIT
    Springer
    DOI: 10.1007/978-3-642-11577-6_33
Wen-Hung Liao1,*, Chien-An Lai1,*
  • 1: National Chengchi University
*Contact email: whliao@cs.nccu.edu.tw, g9202@cs.nccu.edu.tw

Abstract

The proliferation of digital cameras has changed the way we create and share photos. Novel forms of photo composition and reproduction have surfaced in recent years. In this paper, we present an automatic caricature generation system using transformative approaches. By combing facial feature detection, image segmentation and image warping/morphing techniques, the system is able to generate stylized caricature using only one reference image. When more than one reference sample are available, the system can either choose the best fit based on shape matching, or synthesize a composite style using polymorph technique. The system can also produce multiple expressions by controlling a subset of MPEG-4 facial animation parameters (FAP). Finally, to enable flexible manipulation of the synthetic caricature, we also investigate issues such as color quantization and raster-to-vector conversion. A major strength of our method is that the synthesized caricature bears a higher degree of resemblance to the real person than traditional component-based approaches.