Share Email Print

Proceedings Paper

Shape from shading-based 3D face shape recovery from single 2D image
Author(s): Fenlan Li; Kexin Xu
Format Member Price Non-Member Price
PDF $14.40 $18.00
cover GOOD NEWS! Your organization subscribes to the SPIE Digital Library. You may be able to download this paper for free. Check Access

Paper Abstract

3D depth information of a face, which does not vary under different exterior conditions, provides an excellent clue to current face recognition (FR) research. Among techniques of 3D shape recovery, shape from shading (SFS) method which can recover from single 2D image is most popular. However, most existing SFS algorithms assume that objects in the scene are Lambertian reflectors and the reflectivity of the object's surface is uniform, which makes it difficult to reconstruct the face shape successfully and accurately, mainly because the face not only has a complex shape but also is composed of parts with different reflecting properties such as cheek and lip. In this paper, an improved algorithm is proposed. The strategy consists in: 1) according to face's shape and albedo symmetry, an albedo free reflectance function is derived; 2) through an eye detector, the 2D face image being measured is aligned to a 3D face model; 3) the iteration equation for computation of the depth value is obtained and the 3D depth information of the model is set as initial depth value of the face shape. Through continuously iteration until convergence, the 3D depth information of the face is achieved lastly; 4) based on the depth values calculated, the pixel-wise albedo value is also obtained. Experiments are carried out on symmetry objects and real face images. Results show that the proposed SFS algorithm can efficiently accomplish face's 3D shape recovery in less time and at less iteration.

Paper Details

Date Published: 27 October 2006
PDF: 7 pages
Proc. SPIE 6047, Fourth International Conference on Photonics and Imaging in Biology and Medicine, 60473S (27 October 2006); doi: 10.1117/12.710741
Show Author Affiliations
Fenlan Li, Tianjin Univ. (China)
Kexin Xu, Tianjin Univ. (China)

Published in SPIE Proceedings Vol. 6047:
Fourth International Conference on Photonics and Imaging in Biology and Medicine
Kexin Xu; Qingming Luo; Da Xing; Alexander V. Priezzhev; Valery V. Tuchin, Editor(s)

© SPIE. Terms of Use
Back to Top