A nonparametric approach to face detection using ranklets

Smeraldi, Proceedings of the 4th International Conference on Audio and Video-based Biometric Person Authentication (AVBPA 03), Guildford, UK, pages 351-359, June 2003

Abstract

Ranklets are multiscale, orientation-selective, nonparametric rank features similar to Haar wavelets, suitable for characterising complex patterns. In this work, we employ a vector of ranklets to encode the appearance of an image frame representing a potential face candidate. Classification is based on density estimation by means of regularised histograms. Our procedure outperforms SNoW, linear and polynomial SVMs (based on independently published results) in face detection experiments over the 24'045 test images in the MIT-CBCL database.

Full paper (PostScript)

Backlinks: Publications