Function INVERSE_SCAT
Package
reconstruction
Short description
Reconstruct a signal from its scattering coefficients
Usage
xt = inverse_scat(S, filters, options, node)
Input
S (cell): The scattering coefficients output by SCAT.
filters (cell): The set of filter banks used to calculate S. Can be
obtained from FILTER_BANK or second output of WAVELET_FACTORY_1D.
options (struct, optional): Specifies different parameters for the
inversion. These are passed on to GRIFFIN_LIM and RICHARDSON_LUCY, so
please see the documentation for these functions for specific
parameter settings.
node (numeric, optional): A vector of the form [m p], where m is the
coefficient to recover and p is its index. If [S,U] is the output of
the SCAT function, this corresponds to estimating U{m+1}.singal{p}.
To recover the original signal, node is therefore [0 1] (Default
[0 1])
Output
xt (numeric): The reconstructed signal.
Description
The scattering transform is inverted recursively, estimating
U{m+1}.signal{p} for each layer, starting with the last and propagating
upwards until U{1}.signal{1} is obtained, which is an estimate of the
original signal.
For the last layer m = l, the Richardson-Lucy deconvolution algorithm is
applied to estimate U{l+1}.signal{p} from S{l+1}.signal{p}. Previous
layers are reconstructed through the Griffin & Lim algorithm, with the
estimate of U{m+1}.signal{p1} are obtained from the estimates of
U{m+2}.signal{p2}, where p2 correspond to the wavelet modulus coeffi-
cients in layer m+1 of the coefficient p1 in layer m -- the former being
the "children" of the latter.
See also
List of all packages