얼굴의 특징점을 이용한 3차원 움직임 추적에 관한 연구

A Study on the 3D Motion Tracking Using the Facial Feature Points

In this paper 3-dimensional feature points with 2-dimensional image of human face are eventually extracted. For the 2-dimensional image, avi file is created as a transfer medium for real-time movie transfer using 2 CCD cameras. The coarse-tᄋ-fine method is applied to minimize matching error when extracting feature-point of a face. By stereo-matching between side and front image, 3-dimensional data are extracted. As utilizing those extracted 3-dimensional data, various 3D graphic package will be connectable. With two usb CCD cameras, 2-dimensional is still received image. With movements of 20 feature points from Side and front image, we extracted 3-dimensional feature-points and enabled 3-dimensional modeling. To minimize errors of 3-dimensional trace and matching error spread, Coarse-to fine method is applied.