Rotation and Scale Invariant Texture Analysis with Tunable Gabor Filter Banks

In this paper, we propose a method that can be used for image texture recognition in the presence of concurrent rotation and scale changes with tunable directional bandpass Gabor filter banks. The method relies on the analysis of the frequency spectra of the image textures, and from which the rotation and scale changes are estimated using a new spectral shift measure. Tunable Gabor filter banks are designed based on the spectral shift measure. Spectral features obtained from applying the tuned Gabor filter bank are used in a novel search strategy to achieve texture recognition. The proposed method is compared with a non-tunable Gabor filter bank and the improvement in recognition performance is demonstrated through the experimental results on 112 Brodatz textures.