Bandpass2 - aleslab/Psychtoolbox-3-aleslab-fork GitHub Wiki

w=Bandpass2(size,fLow,fHigh) returns a "window", i.e. a matrix meant to be
used as a band-pass filter. The matrix size is mxn if "size" is [m,n],
and nxn if "size" is n. The matrix elements represent gain at each freq,
uniformly spaced from about -1 to 1 of Nyquist frequency (see
FREQSPACE). fLow and fHigh are the radial cut-off frequencies on this
scale. The filter has gain 1 in the frequency interval [fLow,fHigh], and
gain 0 outside it. Add EPS to create complementary filters that add to
1, e.g.
Bandpass2(n,0,f)+Bandpass2(n,f+eps,1)==Bandpass2(n,0,1)
For circular symmetry make fHigh<=1. Setting fLow=0 and fHigh=Inf will
produce an all-pass filter. Here's a typical use, to produce bandpass
noise:
noise=randn(n,n);
filter=Bandpass2(n,fLow/fNyquist,fHigh/fNyquist);
if any(any(filter~=1)) % skip all-pass filter
ft=filter.*fftshift(fft2(noise));
noise=real(ifft2(ifftshift(ft)));
end
Also see OrientationBandpass, Bandpass, FREQSPACE.