Measuring 3-D shapes of a human face using photometric stereo method with color light sources and slit patterns

Abstract

We propose a new method for measuring 3-D facial shapes. This method uses two point light sources with red filters and a slit pattern projector with a blue filter. The time required for sampling data can be reduced by switching the color light rays at high speed, and hence a series of 3-D shapes of a human face at each instant can be measured. First, two facial images projected by both the red light rays and the blue light stripes are taken by the video camera, and these images are sepa- rated into two shading images and one slit image using the color infor- mation. Next, the normal vectors at points on the face are computed from the two shading images using the photometric stereo method. The 3-D coordinates of the points on the light stripes are measured from the slit image using the stereo vision algorithm. Finally, all the computed results are blended and the 3-D shape of the face is computed. Experimental results are shown to demonstrate the feasibility of the proposed method. © 2003 Society of Photo-Optical Instrumentation Engineers.