In this paper, rotation invariance and the influence of rotation interpolation methods on texture recognition using\nseveral local binary patterns (LBP) variants are investigated.\nWe show that the choice of interpolation method when rotating textures greatly influences the recognition capability.\nLanczos 3 and B-spline interpolation are comparable to rotating the textures prior to image acquisition, whereas the\nrecognition capability is significantly and increasingly lower for the frequently used third order cubic, linear and nearest\nneighbour interpolation. We also show that including generated rotations of the texture samples in the training data\nimproves the classification accuracies. For many of the descriptors, this strategy compensates for the shortcomings of\nthe poorer interpolation methods to such a degree that the choice of interpolation method only has a minor impact.\nTo enable an appropriate and fair comparison, a new texture dataset is introduced which contains hardware and\ninterpolated rotations of 25 texture classes. Two new LBP variants are also presented, combining the advantages of\nlocal ternary patterns and Fourier features for rotation invariance.
Loading....