Abstract
A binary shape representation called the Hilbert Morphological Skeleton Transform (
HMST) is introduced. This representation combines the Morphological Skeleton Transform (
MST) with the clustering capabilities of the Hilbert transform. The
HMST preserves the skeleton properties including information preservation, progressive visualization and compact representation. Then, an object recognition algorithm, the Hilbert Skeleton Matching Algorithm (
HSMA), is introduced. This algorithm performs a single sweep over the
HMSTs and renders the similarity between them as a distance measure. Testing the
HSMA against the Skeleton Matching algorithm (
SMA) and invariant moments revealed that the
HSMA algorithm achieves slightly better object recognition rates while substantially reducing the complexity. In an experiment of 14,400 shape matches, the
HSMA achieved a 90.36% recognition rate as opposed to 89.76% for the
SMA and 89.49% for invariant moments. On the other hand, the
HSMA improved the
SMA processing more than 40%.
