next up previous
Next: Multiscale medial axis Up: Survey of Current Techniques Previous: Summary of low-level operators

Shape Recognition

In order to utilize or to extract the shape information of objects in an image, a method for representing shape is needed. One such representation, known as the medial axis, was introduced by Blum to describe the shapes of various biological structures [10]. A recent refinement of the medial axis method, the multi-scale medial axis (MMA) technique, was proposed by Coggins and Pizer [20] [21]. Both methods represent the shape of an object by applying a transform that skeletonizes the object. In formal terms, the skeleton (or medial axis) is defined as the set of all points in an object that have at least two points that are equidistant to the object's boundary. The medial axis provides a compact way of representing shape because regions become reduced to curves that follow the general shape of an object. This is useful in applications such as character recognition, in which shape information is more important than area or volume information.





Ramani Pichumani
Mon Jul 7 10:34:23 PDT 1997