Methods from computer vision and scale-space theory are applied to the study of sea-ice motion in Antarctica. The input data is a sequence of daily images of the continent, obtained from scatterometer data and processed with a resolution enhancing algorithm. The information contained in these images can be studied at different scales when the appropriate filters are applied. Large scales omit detail and smoothen local variations of intensity while smaller scales show much detail and local variation. When motion is studied through different scales, different patterns might be observed. We assume that all the information coded in these images is the radar backscatter, and that it is closely coupled with advection. The Optical Flow method is used to obtain a dense vector field representing sea-ice motion, the method’s limitations are overcome by adding second order constraints to the main equation and through the use of large neighborhoods to normalize the direction of flow. Validation of results has been done to the extent possible, taking into account that there is practically no ground-truth data available for Antarctica in the form of buoy-data. Sea-ice motion results are displayed along available ocean surface wind data, observing a clear consistency along the ocean-ice border. The results are compared to existing studies applying wavelets and it is shown that differences can be explained by the fact that each method is observing motion at a different scale.