Rotation invariant texture classification with dominant orientation estimation based on Gabor filters

This paper proposes a new method to estimate the dominant orientations of textures using Gabor filters, along with its modified version to fit the multi-orientation cases. The discrete wavelet transform is used as a feature extraction tool and nearest neighbor method as a classifier. The experiments are carried out on both Brodatz database and CUReT database. The result shows that our method is effective for the classification of rotated textures and low in computational cost.