EFFICIENT FEATURE-BASED CONTOUR EXTRACTION.
- James R.
Extraction of contours in binary images is an important element of object recognition. This paper discusses a more efficient approach to contour representation and generation. This approach defines a bounding polygon as defined by its vertices rather than by all enclosing pixels, which in itself is an effective representation. These corners can be identified by convolution of the image with a 3 x 3 filter. When these corners are organized by their connecting orientation, identified by the convolution, and type, inside or outside, connectivity characteristics can be articulated to highly constrain the task of sorting the vertices into ordered boundary lists. The search for the next bounding polygon vertex is reduced to a one dimensional minimum distance search rather than the standard, more intensive two dimensional nearest Euclidean neighbor search.
- Research Organization:
- Los Alamos National Laboratory
- Sponsoring Organization:
- DOE
- OSTI ID:
- 975663
- Report Number(s):
- LA-UR-01-4205
- Country of Publication:
- United States
- Language:
- English
Similar Records
Wavelet Representation of Contour Sets
A discrete dynamic contour model