Contour Surface Encoding
After the contour relaxation, each uniform region in the sequence is bounded
by a contour surface. Our goal is to encode the surface efficiently so that
high compression ratio can be achieved. According to our experience in
working with individual frame using the segmentation approach, the majority
of bits required to encode the sequence information is consumed to encode
the contour map.
We plan to use a combination of some efficient surface fitting methods such
as the bicubic splines to encode the contour surface so that the contour
surface degradation is minimized.