Pseudo-tensor product interpolation and blending with families of univariate schemes Original Research Article

Tensor product and blended interpolants have traditionally been constructed from two commuting univariate interpolation projectors, one in the x and the other in the y-direction. Here we construct analogous interpolants from two families of noncommuting projectors in the two directions. This makes it possible to use interpolation schemes adapted to the data. In particular, we show how weighted splines may be used to form such interpolants of rapidly varying data.