Rotation- and scale-invariant texture classification using slide matching of the gabor feature

This paper proposes a set of efficient algorithms for rotation- and scale-invariant texture classification. This set is based on the well established Gabor feature. A circular sum of the Gabor feature elements belonging to the same scale is proposed to reduce the effect of rotation, while a slide matching of augmented scales is proposed to address the effect of scaling. The resulting feature vector is more compact, and the distance measure requires less computation. Experimental results indicate that this algorithm is effective for classifying texture images under different scales and rotations.