The univariate B-spline approximation is widely used in many practical curve fitting. In this paper we apply the univariate B-spline with the blending-function method for surface fitting and compare with other schemes such as the product scheme (Gregory``s square with zero twists), and the piecewise quadratic triangular scheme. The graphical and numerical comparisons show that the univariate B-spline with the blending-function method gives the closest approximated surface to the test surfaces from which the data are taken. This is also not inferior to the bivariate B-spline least square approximation in the same sense.