Having determined extended boundaries by grouping together pixels which define discontinuities in the 2D plane, it is necessary to produce a terse and useful representation of these boundaries for storage and matching. The model representation may be a function of the particular algorithm used to determine the boundaries; for example, chain codes or curves (considered below) are a frequent consequence of an edge tracking mechanism. Parameter based representations arise naturally from Hough transformation. The method of representation should be succinct, but should also optimise the search process; there may be justification for storage of redundant information to speed up the matching process.
Comments to: Sarah Price at ICBL.
(Last update: 22th April, 1996)