Rotation Invariant Local Frequency Descriptors for Texture Classification

Abstract

This paper presents a novel rotation invariant method for texture classification based on local frequency components. The local frequency components are computed by applying 1-D Fourier transform on a neighboring function defined on a circle of radius R at each pixel. We observed that the low frequency components are the major constituents of the circular functions and can effectively represent textures. Three sets of features are extracted from the low frequency components, two based on the phase and one based on the magnitude. The proposed features are invariant to rotation and linear changes of illumination. Moreover, by using low frequency components, the proposed features are very robust to noise. While the proposed method uses a relatively small number of features, it outperforms state-of-the-art methods in three well-known datasets: Brodatz, Outex, and CUReT. In addition, the proposed method is very robust to noise and can remarkably improve the classification accuracy especially in the presence of high levels of noise.

3D Solid Texture Classification Using Locally-Oriented Wavelet Transforms

“…Tables I and II summarize the results obtained for each subset, where the best performance achieved by each alignment method is shown. Moreover, Table I contains the performance of the methods obtained in [59] on the Normal subset. a) Fourier dataset: The average accuracy of a random multi-class classifier is 1/l, where l is the number of classes.…”

“…For the classification task, Maani et al used this database to compare existing 3DST descriptors in [55]. These were: 3D GLCM [25], 3D LBP [56], second orientation pyramid (SOP) filtering [57], [58] and a novel approach based on the local frequency descriptor (LFD) [59]. Unfortunately, only the Normal subset of the Fourier dataset was used and can thus be compared to our approach.…”

Texture classification using joint statistical representation in space-frequency domain with local quantized patterns

“…We can see that the proposed method consistently outperforms the state of the art for all three test settings, demonstrating its robustness to rotation and illumination changes. By exploring the joint statistics of local quantized patterns in the space-frequency domain, our method outperforms the space-based multi-scale CLBP and [15] and the frequency-based LFD [11]. Table II further reveals that the gradient and LOG channels can provide complementary information and combing them leads to the best performance.…”

“…Then, we keep the first K (K ≤ P/2 + 1) magnitude components in view of the symmetric property of Fourier coefficients. This step also gives the rotation invariance and preserves the main energy of the texture [11]. Finally, these magnitude components are L-2 normalized to be robust to illumination changes.…”

Texture classification using rotation invariant models on integrated local binary pattern and Zernike moments

“…Many researchers developed LBP methods based on Ojala's idea. For example, Zhao et al [21], Maani et al [22], and Ahonen et al [23] respectively improved the LBP method using frequency domain analysis methods. Mäenpää [24] pointed out that texture can be regarded as a two-dimensional phenomenon characterized by two orthogonal properties: patterns and the strength of the patterns, and these two measures are supplementary to each other in a very useful way.…”