Feature-based Digital Head Reconstruction

• Main Authors: Ying-Lung Huang, 黃盈倫
• Other Authors: Jing-Jing Fang
• Format: Others
• Language: zh-TW
• Published: 2004
• Online Access: http://ndltd.ncl.edu.tw/handle/04393248711286709298

碩士 === 國立成功大學 === 機械工程學系碩博士班 === 92 === 　　The research invoking body scanner to reconstruct human digital model has carried on for a few decades. Digital head reconstruction is also one of its important research topics. Due to the features of human head is rather complex and changeable, the issue of how to preserve the original features of the digital head and how to simplify it for the purpose of generating coherence facial expression, become crucial issues. The main problem of the issue starts from the human body scanner, the extracted cloud data are huge and unstructured. If we manually pinpoint the feature points, it may lack of uniqueness from the previous selection. Therefore we develop an automatic feature extraction system, in order to reconstruct the digital head from those feature points and feature lines. The outcomes reveal both of simplifying scanning data and preserving the head’s geometric features simultaneously. Based on it, we are able to apply to multi-media image transmission or real like emulation in computer animation. 　　In this thesis, we introduce the mathematically definitions of the feature points which are mostly defined in ISO/IEC/JTCI/SC29/WG11N4030 MPEG-4. By invoking computer algorithms to extract features on a scanning head, we are able to re-construct the digital head automatically. In addition, for the purpose of solving the problem of lacking image feature data, we developed textural mapping technique to match both pictures and geometric head. A real-like 3D digital head on screen is possible.