Advances in Computer Science and Engineering Volume 6, Number 1, 2011, Pages 1-32 Published Online: February 22, 2011 This paper is available online at http://pphmj.com/journals/acse.htm © 2011 Pushpa Publishing House : phrases and Keywords 3D face recognition, shape-from-shading, global, local models, albedo estimation, expectation maximization. Communicated by Tarek M. Sobh Received April 28, 2010; Revised September 23, 2010 A NEW APPROACH IMPROVING THE QUALITY OF 3D FACE RECOGNITION AHMAD RADMAN, ABDULSALAM ALKHOLIDI and HABIB HAMAM ∗ Electrical Engineering Department Faculty of Engineering Sanaà University Republic of Yemen e-mail: ahmed_radmank@yahoo.com abdulsalam.alkholidi@gmail.com ∗ Faculty of Engineering University of Moncton NB, Canada e-mail: habib.hamam@umoncton.ca Abstract Face recognition is one of the most intensively studied topics in the field of computer vision and pattern recognition. In this paper, two statistical models of facial shadow and shape, embedded within a shape-from- shading (SFS) algorithm, are used to reconstruct a 3D facial shape and albedo from 2D image. We developed an iterative process in a statistical setting using Hidden Markov Model (HMM) in Expectation Maximization (EM) algorithm and the Principal Components Analysis (PCA) algorithm to model the likelihood for the shadow regions and surface normal. Our method improves the statistical model results by applying some constraints with feedback error soft correction to decreasing the error that results from non-Lambertian, shadowing, non-symmetric and non-