A Fast, Memory-Efficient Alpha-Tree Algorithm using Flooding and Tree Size EstimationYou, J., Trager, S. & Wilkinson, M. H. F., 11-Mar-2019, (Accepted/In press).
Research output: Contribution to conference › Paper › Academic
The alpha-tree represents an image as hierarchical set of alpha-connected components. Computation of alpha-trees suffers from high computational and memory requirements compared with similar component tree algorithms such as max-tree. Here we introduce a novel alpha-tree algorithm using 1) a flooding algorithm for computational efficiency and 2) tree size estimation (TSE) for memory efficiency. In TSE, an exponential decay model was fitted to normalized tree sizes as a function of the normalized root mean squared deviation (NRMSD) of edge-dissimilarity distributions, and the model was used to estimate the optimum memory allocation size for alpha-tree construction. An experiment on 1256 images shows that our algorithm runs 2.27 times faster than Ouzounis and Soille's thanks to the flooding algorithm, and TSE reduced the average memory allocation of the proposed algorithm by 40.4%, eliminating unused allocated memory by 86.0% with a negligible computational cost.
|Publication status||Accepted/In press - 11-Mar-2019|
|Event||International Symposium on Mathematical Morphology - Saarland University, Saarbrücken, Germany|
Duration: 8-Jul-2019 → 10-Jul-2019
Conference number: 14
|Conference||International Symposium on Mathematical Morphology|
|Period||08/07/2019 → 10/07/2019|
International Symposium on Mathematical Morphology
08/07/2019 → 10/07/2019Saarbrücken, Germany
No data available