Fast Algorithms for the Computation of Ranklets
School of Electronic Engineering and Computer Science Research Reports;EECSRR-09-01 - January 2009
MetadataShow full item record
Ranklets are orientation selective rank features with applications to tracking, face detection, texture and medical imaging. We introduce efficient algorithms that reduce their computational complexity from O(N logN) to O(!N + k), where N is the area of the filter. Timing tests show a speedup of one order of magnitude for typical usage, which should make Ranklets attractive for real-time applications.