Rotation and scale invariant texture features using discrete wavelet packet transform

Abstract

Novel rotation and scale invariant features are proposed in this paper using discrete wavelet packet transform. The classification performance is tested on a set of 15 Brodatz textures rotated in 12 directions and for five scales across an octave. The classification performance for different wavelet filter banks for the proposed rotation and scale invariant features is tested. An application of these features for script identification is illustrated.