www.Ranklets.net
Ranklets are a complete family of multiscale rank features characterized by Haar-wavelet style orientation selectivity. Closely related to the Wilcoxon rank sum test, ranklets are computationally efficient and admit an intuitive interpretation in terms of pairwise comparisons of pixel intensity values.
In the following you will find information on the definition of ranklets, their computation and sample pattern recognition applications.
- Definition of Ranklets
- Orientation selectivity
- Relation to the Wilcoxon statistics (and efficient computation)
- Completeness
- Applications
Ranklets were originally introduced in:
- Smeraldi: Ranklets: orientation selective non-parametric features applied to face detection, 2002 International Conference on Pattern Recognition, Quebec City, QC, Canada, 2002, pp. 379-382 vol.3, doi: 10.1109/ICPR.2002.1047924 (paper)
For more on applications and recent developments, check out the Bibliography.
Software can be downloaded from this page.