We describe a method of bulge-to-disk decomposition which utilizes a f
ull two-dimensional image of a galaxy. It provides much better estimat
es of the structural parameters for each component than the widely use
d conventional procedure which involves a chi(2) decomposition on elli
ptically averaged profiles. The performances of the conventional one-d
imensional method and the present two-dimensional technique are compar
ed by using a set of realistic model galaxy images. We show that the c
onventional one-dimensional decomposition procedure is subject to stro
ng systematic errors associated with the derivation of the ellipticall
y averaged radial surface brightness profiles; these errors come from
the different intrinsic shapes of disk and bulge, combined with the ef
fects of viewing inclination. The present two-dimensional approach suc
cessfully recovers the intrinsic parameters for a wide range of B/D ra
tios at all inclinations.