There have been several recent studies outlining attempts to extract Arctic sea ice motion from Seasat synthetic aperture radar imagery (Hall and Rothrock, 1981; Curlander et al., 1985; Fily and Rothrock, 1986) These studies all highlighted the need to automate, or at least minimize, the amount of operator intervention involved in the procedure. This paper presents the results of a fully automated system to extract ice motion from an image pair. The problem of finding a field of displacements between two sequential images is posed as an estimation problem, and the system is shown to be an optimal solution to this problem. The performance of the system is shown to be superior to ice motion systems that use only cross correlation. The shortcomings of the system, and of area correlation in general, are outlined, and ways of overcoming them are suggested. ¿ American Geophysical Union 1988