The rotational de la Vallee Poussin kernel is defined by
\[ K(t) = \frac{B(\frac32,\frac12)}{B(\frac32,\kappa+\frac12)}\,t^{2\kappa}\]
for \(t\in[0,1]\), where \(B\) denotes the Beta function. The de la Vallee Poussin kernel additionaly has the unique property that for a given halfwidth it can be described exactly by a finite number of Fourier coefficients. This kernel is recommended for Texture analysis as it is always positive in orientation space and there is no truncation error in Fourier space.
Hence we can define the de la Vallee Poussin kernel \(\psi_{\kappa}\) depending on a parameter \(\kappa \in \mathbb N \setminus \{0\}\) by its finite Chebyshev expansion
\[ \psi_{\kappa}(t) = \frac{(\kappa+1)\,2^{2\kappa-1}}{\binom{2\kappa-1}{\kappa}} \, t^{2\kappa} = \binom{2\kappa+1}{\kappa}^{-1} \, \sum\limits_{n=0}^{\kappa} (2n+1)\,\binom{2\kappa+1}{\kappa-n} \, \mathcal U_{2n}(t)\].
Syntax
Input
kappa | kernel parameter |
Options
halfwidth | angle at which the kernel function has reduced to half its peak value |
bandwidth | harmonic degree |