Using Workbench Command

INTERPOLATE BETWEEN DISJOINT DATA
   wb_command -metric-interpolate-gaps
      <surface> - the surface to compute on
      <distance> - limit for how large of a gap to interpolate, in mm
      <metric-out> - output - the output metric

      [-data-roi] - specify a single-map roi, locations outside it will have
         their input data ignored, and will have zeros for the output values
         <roi-metric> - metric file, positive values denote vertices that have
            usable data

      [-corrected-areas] - vertex areas to use instead of computing them from
         the surface
         <area-metric> - the corrected vertex areas, as a metric

      [-dilate-exponent] - specify a different exponent for within-map dilation
         <exponent> - exponent 'n' to use in (area / (distance ^ n)) as the
            weighting function (default 6)

      [-dilate-nearest] - use the nearest good value instead of a weighted
         average for within-map dilation

      [-dilate-linear] - fill in values with linear interpolation along
         strongest gradient for within-map dilation

      [-metric] - repeatable - specify an input metric
         <metric-in> - the input data
         <bad-vertex-roi> - metric file, positive values denote vertices of the
            input data that don't have good data

      Each input metric is dilated independently, and then the results are
      blended based on each vertices' distance to the closest vertex that has
      good data in each input.  When the closest vertex for an input is blocked
      by a good vertex from a different input, the blocked input is not used
      for that vertex.  When only two inputs are used at a vertex, the blending
      between dilated inputs is a linear interpolation based on the closest
      distances.

      Other options work as described in -metric-dilate.