Face modeling using grid light and feature point extraction

Abstract

In this paper, an algorithm for extracting three-dimensional shape of human face (Face Modeling) from 2D images using grid light is proposed. The grid light is illuminated by common white light instead of laser light in order to protect the human eyes or skin and reduce cost. A simple and uncoded grid pattern is projected on human face to solve the problem of correspondence between a pair of stereo images. The grid stripes are extracted and thinned by applying first smoothing and then a marker watershed segmentation algorithm. For the sake of providing more details for facial model, feature point extraction is introduced. The set of matched feature points will be added to the set of matched points. The final set of matched points is further used to calculate three-dimensional depth information of face. Experimental results have shown the feasibility of the proposed method. © Springer-Verlag Berlin Heidelberg 2005.