Abstract
Distance transforms on the face-centered cubic (fee) grid and the body-centered cubic (bcc) grid are examined. Since the voxels on the fee and bee grids are better approximations of a Euclidean ball than the cube, the distance transforms (DTs) on these grids can be less rotation dependent than those in Z(3), which is a desirable feature. Optimal (according to the error function) weights are calculated and integer approximations of these weights are found. Also, the two-dimensional city block distance is generalized to the fee and bee grids by considering a unit distance between gridpoints whose corresponding voxels share a face. A method to compute the DTs is presented. The results are evaluated both theoretically and by actually computing some DTs. (c) 2005 Elsevier Inc. All rights reserved.
| Original language | English |
|---|---|
| Pages (from-to) | 294-311 |
| Number of pages | 18 |
| Journal | Computer Vision and Image Understanding |
| Volume | 100 |
| Issue number | 3 |
| DOIs | |
| Publication status | Published - 2005 |
Keywords
- distance transform
- fcc
- bcc
- non-cubic voxels
Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver