SoGaborFilteringProcessing2d Class Reference
[Frequency Domain]

SoGaborFilteringProcessing2d image filter More...

#include <ImageViz/Engines/ImageFiltering/FrequencyDomain/SoGaborFilteringProcessing2d.h>

Inheritance diagram for SoGaborFilteringProcessing2d:

Public Member Functions
	SoGaborFilteringProcessing2d ()
Public Attributes
SoSFImageDataAdapter	inImage
SoSFInt32	frequencyNumber
SoSFInt32	angleNumber
SoImageVizEngineOutput < SoSFImageDataAdapter, SoImageDataAdapter * >	outImage

Detailed Description

SoGaborFilteringProcessing2d image filter

The SoGaborFilteringProcessing2d engine makes filtered images thanks to Gabor filters, which are well known for localisation in frequency and space. All these filters are made in the Fourier Space (frequency space). Gabor filters recover this space.

For a number of angles $N_\theta$ and a number of frequences $N_\omega$ , we can define $N_\theta\times N_\omega$ filters of Gabor :

for the angles $theta=\frac{n_\theta\times\pi}{N_\theta},0\leq n_\theta<N_\theta$ ,

and the frequencies $F=\frac{F_n}{2^{n_\omega}},0\leq n_\omega<N_\omega$ . being the Nyquist frequency.

The frequency responses are:

$G(i,j)=e^{-2\times pi^2 \times \sigma^2\times (u-F)^2\times \lambda^2 + v^2}$

where , $u=\widetilde{i}\times cos(\theta)+\widetilde{j}\times sin(\theta)$ , $v=-\widetilde{i}\times sin(\theta)+\widetilde{j}\times cos(\theta)$ , $\widetilde{i}$ and $\widetilde{j}$ being the centered and normalized coordinates.

$\sigma=\frac{\sqrt{(log(2))/2}}{\pi \times F \times tan(\pi / (2\times N_\theta))}, ~ \lambda=\frac{2^B+1}{2^B-1}\times tan \Big(\frac{\pi}{2\times N_\theta}\Big) ~~ \mbox{where B is given}.$