In this thesis, a method for fitting real object surfaces to mathematical surface is presented. Surface fitting is the process which constructs the concise representation to model the considerably large number of input points. In this paper, instead of recursive subdivision of input points, sequential subdivision is proposed. This method processes input data only once. Hence sequential subdivision method processes data more faster than recursive subdivision. For the convenience of sequential subdivision, surface points are classified into two types, to say, boundary points and inner points. Firstly, the allowable polygonal approximation is performed upon the boundary points. Then the polyhedral division of inner points is followed and the knot points are generated as a result. A special test scheme is proposed. This scheme simplifies acceptance test and time efficiency is obtained again as a result. The method has been implemented using a piecewise parametric bicubic B-spline surface. The comparision of original surface with mathematical surface is performed by displaying two surfaces. This method is simple in concept. This method obtains time efficiency in two points, sequential data processing and test simplification. And more data reduction is acquired by adopting B-spline surface than Bezier surface.