Separating ice-water composites and computing floe size distributions

In this paper we describe an image processing technique that separates ice-water composites and determines the floe size distribution and concentration from SAR polar imagery. At marginal ice zones (MIZs), ice dynamics are active, providing important information for climatic indications and polar navigation. Hence, the identification of the ice-water composites and floe size distribution in these areas is of importance to geophysicists and polar navigator. We have designed a multi phase technique that combines Gaussian approximation, Laplacian measurement and structural modeling to obtain ice-water segmentation and floe extraction. With this technique we compute floe size distribution, ice-water composites, and coverage statistics. We have successfully tested our technique on a database of SAR MIZ imagery and established floe distribution and ice coverage maps for the corresponding areas.