An objective technique is developed to calculate advective surface velocities from sequential images. A pattern matching method is used with the identification of maximum cross correlation (MCC) between a template window in the first image and search areas in the second image. We examined the limitations of the MCC method in two cases: (1) eddy size (LE) much smaller than the radius of deformation (Lp), and (2) LE≥Ls. In the first case an eddy is regarded as a particle. Maximum detectable time period is estimated to be about 1 day. For the second case we developed the MCC method to detect a rotational motion. We compared velocity fields drived from the method with a real velocity field in a numerical analysis of tracer and quasi-geostrophic eddy fields. The optimal template size is about the eddy diameter which spans between the maximum velocity points. We also applied the MCC method to infrared images in the Kuroshio-Oyashio confluence zone. ¿ American Geophysical Union 1989