This paper presents a photometric stereo algorithm that reconstructs o
bject shapes from multiple images, in which given 3D surfaces are appr
oximated by Legendre polynomials and the relationships between the giv
en surface and its derivatives are represented in matrix forms in term
s of a polynomial coefficient vector. The reflectance map is linearize
d and the cost function expressed in quadratic matrix form in terms of
the polynomial coefficient vector is minimized. The relative depth an
d its derivatives are obtained by updating them iteratively. Computer
simulation with various noiseless/noisy sets of test images shows that
the performance of the presented two-image photometric stereo algorit
hm is comparable to that of the conventional methods in terms of three
different error measures: brightness error, orientation error and hei
ght error. Also the performance comparison of the presented and conven
tional three-image photometric stereo algorithms for the noiseless/noi
sy sets of images is shown. (C) 1998 Pattern Recognition Society. Publ
ished by Elsevier Science Ltd. All rights reserved.