The WaveletDenoiseX
node performs noise removal using the fast
wavelet transform[1,2] on a matrix of profiles.
Here are the node arguments:
- Input ports:
X: the raw data matrix. Each row is a profile, (e.g. infrared
spectrum) which will be wavelet transformed.
- Input arguments:
- method: string indicating the different denoising methods
available:
- Visu: A soft thresholding method where the threshold is
- SURE: The Stein's Unbiased Risk Estimate (REF) which is based on
hard thresholding
- Hybrid: Uses sometimes soft thresholding with the Visu threshold
and other times hard thresholding with the SURE threshold
- MinMax: Uses an optimal threshold from estimation of the noise level
- MAD: Median Absolute Deviation, a threshold for each scale is
followed by a soft thresholding
- wavelet: string with name of the wavelet chosen
- par: Integer to indicate the number of vanishing moments
of the wavelet
- L: scalar indicating the coarsest level used for the
reconstruction. The default is 0 containing the coarsest level
possible.
- Output ports:
- Wh: a matrix containing the denoised wavelet coefficient vector
for each profile (e.g. infrared spectrum) in each row
- Xh: the reconstructed data matrix. Each row is a profile,
(e.g. infrared spectrum) which contains the denoised row/spectrum.
SEE ALSO: WaveletDenoiseW
which is very similar to this
function, but differs in the way the the data to be denoised are
presented to the denoising methods. In WaveletDenoiseW
a matrix
with wavelet coefficients from a previous wavelet transform is
presented to the node.
Source: This node is using tools from the WavelLab toolbox which
can be downloaded from http://www-stat.stanford.edu/~wavelab/
Bjørn Kåre Alsberg
2006-04-06