Wavelet denoising#

Install the optional PyWavelets dependency with pip install "spectrochempy[wavelets]", then apply Filter along a spectral dimension.

import spectrochempy as scp

filtered = scp.Filter(method="wavelet").transform(spectrum, dim="x")

By default, the filter uses Symlet-4, symmetric extension, and up to three decomposition levels. It estimates one soft threshold per spectrum from the median absolute deviation of the finest detail coefficients. Use wavelet, wavelet_level, or wavelet_threshold to change these settings. An explicit threshold uses the signal units.

The transform operates in sample space and requires real, finite, unmasked data. The result retains the input shape, coordinates, and units. The processing history records the effective settings.

The smoothing comparison shows the effects on noise, peak position, height, and width.