MIN_CURVE_SURF vs. TRI_SURF - Problem

[<asowter(at)synopticsga.freeserve.co.uk>]

Another cry for help!  I'm a regular user of IDL and am using it to process
RS data.

I have some large regular arrays that have holes in and have tried to
interpolate across them but I get either bad results or the process is too
slow!  TRI_SURF is fastest but gives horrible striping effects with large
variations across larger holes - this is also the case for TRIGRID.  The
best by far is MIN_CURVE_SURF (KRIG2D is really slow) but this tends to slow
down my program to the extent that it takes 3x longer to complete than with
TRI_SURF or TRI_GRID!!!

The call I'm using is:

    result=min_curve_surf(z,x,y,gs=[1,1],bounds=[0,0,xdim0,ydim0])

Z, X and Y contain the value and ordinates of the good (regularly spaced)
pixels within the BOUNDS box.

Any suggestions/alternatives would be much appreciated.

Thanks in advance

Andy