Core-line tracing for fuzzy image subsets
Pattern Recognition Letters
Hi-index | 0.14 |
This paper newly proposes and discusses a maximal square moving (MSM) algorithm for tracing the core lines or medial lines of objects in a digitized binary picture, by which structure-descriptive core-line representations are directly obtained. The MSM algorithm could be considered to be an extension of the medial axis transform (MAT) definition to discrete images since maximal squares are sub-stituted for maximal disks, but it is, in essence, distinct from the con-ventional methods since it is based on what is called an input-time tracing principle. The algorithm is fast and requires no image memory, and the connectivity of the core lines so derived is completely guaranteed in a sense of maximal square neighborhood relations. Furthermore, the original objects can be exactly reconstructed from their core lines derived.