2 Identifying Ice Floes and Computing Ice Floe Distributions in SAR Images
Fig. S. Preservation of size and
shape by the restricted growing
concept. Floes denoted as white
pixels regaining original sizes
and shapes superimposed on the
original ERS-1 SAR image taken
on August 21,1991 at 75.0 oN,
145.3°W (Copyright ESA)
17
ly. Comparing the two images, one can see that separation has been established among
core objects while fuzziness is eliminated from skin objects. Figures 4e and 4f show the
morphologically cleaned core and skin images, respectively. Note that most noise
effects such as peninsulas, lakes, inlets, and islands have been removed. Figure 4g shows
the grown image. Note that the ice floes have been fully restored to their original shapes
and sizes, and separation has been established. Figure 4h shows the floe boundaries
superimposed on the original image. These images exemplify the difference between
the core and skin images, the effects of morphological cleaning, and the relationships
among the grown final output, the core image, and the skin image.
Our primary objective in sea ice floe identification is to establish separation among
ice floes while preserving the original sizes and shapes. Figure 5 illustrates the extraction of floes that are separated and whose size and shape is preserved. The original
image is an ERS-1 sea ice image taken in August of 1991 near the marginal ice zone at
75.03 ON, 145.31 oW. Ice floes are denoted as white pixels superimposed onto the original
image. The grayish regions are water and background-matrix - a mixture of ice and
water. Note that the size and shape of each floe are fully restored while separation
remains established. For example, the large piece of ice floe in the center has been disconnected from its touching neighbors by destroying about a dozen bridges around the
outline of the floe.
2.4
Ice Floe Algorithm and Results
In the following, we present in detail an algorithm or series of procedures for determining floe size distribution in SAR sea ice imagery. Our technique consists of six
stages: (1) image enhancement, (2) image segmentation, (3) floe extraction, (4) floe fil-
Fig. S. Preservation of size and
shape by the restricted growing
concept. Floes denoted as white
pixels regaining original sizes
and shapes superimposed on the
original ERS-1 SAR image taken
on August 21,1991 at 75.0 oN,
145.3°W (Copyright ESA)
17
ly. Comparing the two images, one can see that separation has been established among
core objects while fuzziness is eliminated from skin objects. Figures 4e and 4f show the
morphologically cleaned core and skin images, respectively. Note that most noise
effects such as peninsulas, lakes, inlets, and islands have been removed. Figure 4g shows
the grown image. Note that the ice floes have been fully restored to their original shapes
and sizes, and separation has been established. Figure 4h shows the floe boundaries
superimposed on the original image. These images exemplify the difference between
the core and skin images, the effects of morphological cleaning, and the relationships
among the grown final output, the core image, and the skin image.
Our primary objective in sea ice floe identification is to establish separation among
ice floes while preserving the original sizes and shapes. Figure 5 illustrates the extraction of floes that are separated and whose size and shape is preserved. The original
image is an ERS-1 sea ice image taken in August of 1991 near the marginal ice zone at
75.03 ON, 145.31 oW. Ice floes are denoted as white pixels superimposed onto the original
image. The grayish regions are water and background-matrix - a mixture of ice and
water. Note that the size and shape of each floe are fully restored while separation
remains established. For example, the large piece of ice floe in the center has been disconnected from its touching neighbors by destroying about a dozen bridges around the
outline of the floe.
2.4
Ice Floe Algorithm and Results
In the following, we present in detail an algorithm or series of procedures for determining floe size distribution in SAR sea ice imagery. Our technique consists of six
stages: (1) image enhancement, (2) image segmentation, (3) floe extraction, (4) floe fil-