ROTATION INVARIANT TEXTURE CLASIFICATION USING GABOR WAVELETS

Abstract

A method of rotation invariant texture classification based on spatial frequency model is developed. Features are derived from the multichannel Gabor filtering method. The classification performance is first tested on a set 1440 samples of 15 Brodatz textures rotated in 12 directions (0 to 165 in steps of 15 degrees). For the 13-class problem reported in [13] we got better classification with our features. The total Brodatz album is tested using the same features. 10752 samples from Brodatz album are classified (each texture rotated in 12 orientations). The percentage correct classification is 84.92 for Brodatz album.