An Automatic Face Modeling Algorithm for Image Sequence

概要

論文の詳細を見る
For encoding audiovisual information, the MPEG-4 Synthetic and Natural Hybrid Coding (SNHC) group is proposing an efficient representation and composition of synthetically and naturally generated information. One of the most important applications of this coding standard is model-based coding which provides a promising solution for face-to-face communication. Most researches focus on the facial image analysis and synthesis, but little information is available on the face model adaptation, which is the most significant part of the model-based coding and determines the quality of the face image reproduction. An adaptation algorithm for a generic 3-D face model to an actual face in a head-and-shoulder is proposed for the audiovisual applications. It performs global scaling to reposition and resizing the wireframe, as well as local adaptation to mimic individual specific appearance. A hierarchical approach based on a priori knowledge is exploited to extract the semantic features in the face, which is closely related to model adaptation, such as head, eyes, nostril, and mouth. With these extracted features, the generic wireframe model can be automatically adapted for any person.
社団法人電子情報通信学会の論文
1998-01-21