Optical flow and scale-space theory applied to sea-ice motion estimation in Antarctica

Sea-ice motion in Antarctica is studied applying methods from computer vision and scale-space theory to a sequence of images obtained from scatterometer data. The proposed method can obtain a dense motion vector field for any specific observation scale. Spatial and temporal scales are used to focus on relevant geophysical features and events. A preprocessing stage involving spatial and temporal filtering at selected scales reduces noise and artifacts produced in the image generation phases, allowing reliable feature extraction and tracking at relevant scales. Optical flow (OF) methods provide a dense estimation of the motion field. The limitations and advantages of this approach are discussed. Optical flow sea-ice motion data are in agreement with sea winds data obtained independently and with very different methodologies. OF results are compared to data from wavelet methods.