Rotation Invariant Texture Classification using Gabor Phase Features

Abstract

A method of rotation invariant texture classification based on Gabor phase features is developed. Features are derived from Gabor filtered texture images. The phase gradient renders useful boundary and region information for simple texture images. The classification performance is tested on a set 1440 samples of 15 Brodatz textures rotated in 12 directions (0 to 165 in steps of 15 degrees). The percentage correct classification is 80.49 for the test samples. The results are encouraging. The experiment is extended for a 45-texture (Brodatz) problem.

Indexing terms:

• Gabor phase features
• Rotation invariance
• Texture classification