“Automatic and accurate mesh fitting based on 3D range scanning data” by Nakano, Nonaka, Maejima and Morishima
Conference:
Type(s):
Title:
- Automatic and accurate mesh fitting based on 3D range scanning data
Presenter(s)/Author(s):
Abstract:
In this paper, we present a generic mesh fitting algorithm based on 3D scanned range data. Generally, 3D range data or 3D digitized data is used to manually create a precise model of face and body shapes that requires considerable amount of time and cost. In the Future Cast System [Morishima, 2005], a generic face model is fitted to the face feature points on a subject’s face which have been extracted automatically only from texture data. The subject’s expression is neutral and 3D range data is taken from the front. With our method, range and texture data are acquired using a range scanner, and then facial feature points are then automatically extracted based on both depth pattern template matching and the computer vision approach. First, the feature points of the nose are detected depending on only depth. Feature points for the mouth and eyes are extracted according to color edge and depth features with high accuracy. After extracting all of the feature points, the 3D face model is generated by fitting the generic mesh to extracted points using the scattered interpolation method. Consequently, our method can automatically create a precise model of face, and dramatically reduce the cost of face modeling with robustness against the lighting condition.
References:
1. Shigeo Morishima, Akinobu Maejima, Shuhei Wemlera, Tamotsu Machida, Masao Takebayashi, Future Cast System, ACM SIGGRAPH Sketch, 2005
