Multiple subarea cross correlations have been computed from visible, advanced very high resolution radiometer images of the Beaufort Sea in an automated technique to detect pack ice motion. The computational method is an application of the well-known matched filter and consists of subdividing the first image into a set of subareas and correlating each reference subarea against the appropriate region of the second image. The correlation functions are processed to determine a set of local maxima which define end point vectors of ice displacements from the subarea centers. Resulting velocities agree well with estimates of ice movement obtained by manually measuring invariant ice feature displacements. The velocity fields show the spatial structures of ice motion over scales of 100--500 km with a high degree of pattern consistency and agreement with simple surface wind driving.