FCONV Fast Convolution/Correlation
y = FCONV(x, h, shape) convolves x and h. A deconvolution mode is also possible.
The correlation, instead of the convolution, can also be computed.
This FFT based convolution should provide the same results as conv2 and convn.
It works with x and h being of any dimensionality. When only one argument is given,
the auto-convolution/correlation is computed.
The accuracy of the conv and xcorr operators depends on the sampling. Peaks should
be describes with at least 5 points underneath.
x = input vector (signal)
h = input vector (filter)
shape = optional shape of the return value
full Returns the full two-dimensional convolution.
same Returns the central part of the convolution of the same size as x.
valid Returns only those parts of the convolution that are computed
without the zero-padded edges. Using this option, y has size
[mx-mh+1,nx-nh+1] when all(size(x) >= size(h)).
deconv Performs an FFT deconvolution.
deconv_iter Performs an iterative deconvolution.
correlation Compute the correlation instead of the convolution (one
of the FFT's is then conjugated).
pad Pads the x signal by replicating its starting/ending values
in order to minimize the convolution side effects.
center Centers the h filter so that convolution does not shift
the x signal.
normalize Normalizes the h filter so that the convolution does not
change the x signal integral.
background Remove the background from the filter h (subtracts the minimal value)
ex: c=fconv(a,b); c=fconv(a,b, 'same pad background center normalize');
See also FCONVN, FXCORR, CONV, CONV2, FILTER, FILTER2, FFT, IFFT
Version: Nov. 27, 2018
(c) E.Farhi, ILL. License: EUPL.