The one-dimensional case can be extended to multiple dimensions through the use of the tensor product spline constructs. A spline subspace is defined for each dimension as discussed in the preceding section. Thereafter a tensor product function in the th-dimensional space can be expressed as:
Tensor product splines provide a straightforward and simple generalization of the one-dimensional spline functions. Multivariate approximation of functions by tensor products has a strong directional bias along lines parallel to the axis direction. Difficult functional features running along different directions require a fine mesh in all the functional space in order to achieve accurate approximation. This increases the computational demands of the problem. Recently, advances in mathematical research on the theory and computational methods of multivariate splines [20] have shown great potential in various applications without the constraints of tensor product constructs.
In the two-dimensional space, the equivalent TPS formulation of (2.73) can be written as:
Where is the ith B-spline of order for the knot
sequence , and the number of knots in the X and
Y direction respectively.
As in the one-dimensional case, the coefficients can be
selected to enforce interpolating or more general conditions.