Zubair Khalid
Academic Webpage
Research School of Engineering College of Engineering and Computer Science The Australian National University
Novel Spherical Harmonic Transform (NSHT)  MATLAB Toolbox
Introduction
NSHT Toolbox is a MATLAB based toolbox for spherical harmonic analysis on the sphere. It implements the novel spherical harmonic transform (NSHT) of a bandlimited signal using the optimal number of samples.
Overview
For the accurate representation of a signal on the sphere bandlimited at degree with degrees of freedom in the spectral domain, we developed a new sampling scheme in [1] that requires on the sphere. Thus, the sampling scheme matches the spectral dimensionality of the signal.
For the sampling scheme, we also developed spherical harmonic transform, referred as novel spherical harmonic transform, which is implemented in this toolbox. The spherical harmonic transforms (forward and inverse) allow the computation of transforms with sufficient accuracy and manageable computational complexity and do not require any precomputation associated with spherical harmonic transforms. The precomputation is only required to determine the optimal placement of samples on the sphere. Once optimal sample positions are determined, the implementation of SHT does not require any precomputation.
If you use NSHT toolbox for harmonic analysis which results in publication, please reference the site (http:www.zubairkhalid.org/nsht) and our related academic paper:
[1] Z. Khalid, R. A. Kennedy and J. D. McEwen, “An OptimalDimensionality Sampling Scheme on the Sphere for Fast Spherical Harmonic Transforms,” IEEE Trans. Signal Process., vol. 62, no. 17, pp. 45974610, September 2014. (IEEE Xplore) (arXiv)
Authors and Affiliations
Zubair Khalid
Research School of Engineering, The Australian National University
Rodney A. Kennedy
Research School of Engineering, The Australian National University
Jason D. McEwen
Mullard Space Science Laboratory, University College London
Download
The source code of the NSHT toolbox can be downloaded from here (version: 0.9b).
We provide this code in the hope that it will be useful, but WITHOUT ANY WARRANTY.
If you notice an error or bug or need support, please write to zubair.khalid@anu.edu.au.
Version History
Current version: 0.9b (15 Aug. 2014), Public release of first beta.
License
NSHT toolbox to compute spherical harmonic transforms and carry out harmonic analysis Copyright (C) 2014 Zubair Khalid
This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with this program. If not, see http:www.gnu.org/licenses.
Installation Instructions
Once you download the toolbox as nsht.zip, unzip the code to the directory ‘nsht’.
The MATLAB working directory or path should be set for directory ‘nsht’ which contains the ‘src’ directory, file ‘nsht_install.m’ and nsht demos.
For working directory: nsht, execute the file ‘nsht_install.m’. The installation creates a mex function for the computation of associated Legendre polynomials, which significantly improves the computation time. The installation also checks that the precomputed optimal sample positions are available in the src/Data directory.
Notes: The bandlimit supported in the current version (0.9b) is . The NSHT is shown to compute sufficiently accurate spherical harmonic transform up to bandlimit . However, the optimal sample positions have been precomputed up to .
Installation Requirements: MATLAB with the support of codegen and mex compiler.
Notes on Installation Requirements: For efficient implementation of spherical harmonic transforms, the associated Legendre functions are computed in C. This computation is implemented using MATLAB codegen that generates a C code and mex function from MATLAB code. The mex function is then used for computation. If MATLAB is not supported by mex compiler, the toolbox can still be used. However, it takes more computation time to compute spherical harmonic transform.
Documentation
There is documentation for the MATLAB functions within the source files, which can be accessed through the help or doc command in MATLAB.
Functions
List of MATLAB functions with brief description is provided here.
The documentation for each function in MATLAB can be obtained using the command: doc <function name>.
List of MATLAB functions
nsht_forward
Computes forward spherical harmonic transform.
nsht_indexed_theta
Determine sample positions along colatutude.
nsht_inverse
Computes inverse spherical harmonic transform.
nsht_legmat
Compute Legendre matrix.
nsht_ordered_theta
Determine the optimal placement of rings along theta.
nsht_plot_sampling
Plots the sampling points on the sphere over the surface of the sphere.
nsht_plot_spectral
Plots the spectral representation of the signal as a surface plot for different degrees and orders.
nsht_plot_sphere
Plots the function on the sphere over the sample points in the proposed sampling scheme.
nsht_sampling_points
Determine the sampling points on the sphere for the sampling scheme proposed in the following paper.
nsht_sharmonic
Compute spherical harmonic over optimal sampling spatial grid.
nsht_sharmonic_g
Compute spherical harmonic over given spatial grid.
nsht_sharmonic_matrix
Computes spherical harmonic of all degrees and orders less than over given spatial grid.
Demos (Examples)
Accuracy Analysis
We randomly generate the complex valued bandlimited test signal with real and imaginary parts uniformly distributed in the interval
over the samples of our sampling scheme. The forward SHT, followed by inverse SHT is applied on the test signal to obtain the reconstructed signal bandlimited at .
We repeat the experiment 10 times and then obtain the average values for the maximum reconstruction error and the mean reconstruction error between the test signal and reconstructed signal . See our paper [1] for
details.
Both and are plotted below for the bandlimit in the range .
Computation Times
We measure the computation times, denoted by and to carry out the proposed inverse and forward transforms, respectively.
The computation times are recorded using NSHT MATLAB Toolbox running on a machine equipped with 2.6 GHz Intel Core i7 processor and 64 GB of RAM.
The computation times and in seconds, averaged over 10 different signals, are plotted below.
