This page provides the demo code for the work

Rotation-Invariant HOG Descriptors using Fourier Analysis in Polar and Spherical Coordinates  

Kun Liu, Henrik Skibbe, Thorsten Schmidt, T. Blein, K. Palme, Thomas Brox, Olaf Ronneberger

International Journal of Computer Vision, Online First, 2013. Publisher's Link

Matlab code for 2D dense description

• Download:
  Successfully tested on Matlab R2012b and Ubuntu 10.04 X86_64. It may require some tiny modifications for a different environment.
  
  
• Requires: Matlab | LIBLINEAR | TAS package and image data (version 0) | export_fig
  
  
• We offer the following functions:
  demoFourierHOG -- visualize the descriptor
  FourierHOG -- the descriptor function
  main -- reproduce the experiment "Fourier HOG_2 + linear SVM" on the aerial image dataset
  
  

Matlab demo code for 3D dense description

• Download:
  Successfully tested on Matlab R2012b and Ubuntu 10.04 X86_64. It may require some tiny modifications for a different environment.
  
  
• We use the spherical tensor derivative implementation from
  Reisert, M. and Burkhardt, H., "Spherical Tensor Calculus for Local Adaptive Filtering",
  in: Tensors in Image Processing and Computer Vision, Springer, 2009.
  
  Another package including spherical tensor derivative can be found at https://bitbucket.org/skibbe/sta-imagetoolbox
  
  
• We offer the following functions:
  demoSH -- demo the Spherical harmonics and function projection/reconstruction
  main -- demo the computation of Spherical Harmonic HOG descriptor (using ST derivative for convolution)
  
  
  
• To visualize the HDF5 files, we recommend the ImageJ hdf5 plugin developped by our group.
  

Matlab/C++ source code for 3D shape description

• Download: (for reference only.)
  
  

If you have any questions or feedbacks, please contact the authors of the paper.