Local Frequency Descriptor (Phase)

Considers a neighboring function defined on a circle of radius R at each pixel:

We define Local Frequency Descriptors (LFD) as the coefficients of the 1D Fourier transform of the neighboring function:

$f_n=\sum_{k=0}^{N-1}t_{k}e^{-2\pi ink/N},$

We experimentally observed that the first two channels consist more than 90% of the energy of the texture signals, so we use only two frequency channels. Three sets of descriptor are extracted: one based on the magnitude, and two based on the phase. The phase based features are constructed from the Histogram of Local Orientations which is formed by qauntizing and accumulating phase information.

Download the Matlab code implementation from here: [code]. Please cite the following reference if you use this code:

R. Maani, S. Kalra, and Y.H. Yang, "Rotation invariant local frequency descriptors for texture classification", IEEE Transactions on Image Processing 22(6), pp. 2409-2419, 2013.