# 2D list interpolation within restricted domain

I have some 2D neuroimaging data. Each {x,y} coordinate is associated with a value. I want to smooth this data, and have been having some success using ListInterpolate. The coordinates do not form a complete rectangle, however. For example:

{ {{1, 2}, -0.7073}, {{1, 3}, -0.6373}, {{2, 2}, -0.5774}, {{2, 3}, -0.2186}, {{3, 2}, -0.5343}, {{3, 3}, -0.08759}, {{4, 1}, -0.5024}, {{4, 2}, -0.5748}, {{4, 3}, 0.18907}, {{5, 2}, -0.3938} }

If I use just this data, I get the error:

"ListInterpolation::inder: The order-1 derivative of 1 is not a tensor of rank 1 with dimensions 2."

If I include a value of 0 for the coordinates without values in the data, The interpolation of the adjacent true values is obviously affected. Is there a way to use a "mask" of some sort to specify which coordinates to interpolate between? The coordinates without values are all on the perimeter of the coordinates with values (that is, there are no "holes" in the data).

I could just add values around the perimeter of my data: pad the data in effect with something like the average of the nearest true values, but wonder whether there is another way.

-
What wrong with just using Interpolation? –  wxffles Jun 13 '13 at 23:39