Wavelet Compression
Wavelet Compression reduces the spectral dimensionality of the current hyperspectral dataset using a one-dimensional wavelet transform applied independently to every pixel spectrum.
Overview
Wavelet Compression reduces the spectral dimensionality of the current hyperspectral dataset using a one-dimensional wavelet transform applied independently to every pixel spectrum.
The tool provides two output modes: Reduced Spectrum and Wavelet Components.
The current working cube is replaced by the selected wavelet output. The source data file is not overwritten automatically.
1. Load a hyperspectral dataset.
2. Open Wavelet Compression from the Preprocessing Center.
4. Select the decomposition level.
5. Choose Reduced Spectrum or Wavelet Components.
6. Select the number of output bands/components.
7. For Reduced Spectrum, choose how the new spectral axis should be represented.
8. Review the live Output Preview.
9. Click Apply Wavelet Compression.
10. Save/export the resulting cube only if you want a new file on disk.
The Wavelet family control selects the mother wavelet used for decomposition.
IDCubePro provides Daubechies (db), Coiflet (coif), and Symlet (sym) families.
db4 is used as the default general-purpose wavelet.
The decomposition level controls how many multiscale decomposition stages are applied.
Higher levels separate progressively broader spectral structure from finer spectral details.
The maximum useful level depends on both the number of spectral bands and the selected wavelet. IDCubePro automatically limits the choices.
Reduced Spectrum reconstructs each spectrum using only the approximation coefficients at the selected decomposition level.
The reconstructed spectrum is smoother because the detail coefficients are suppressed, and it is then resampled to the requested number of output channels.
Use this mode when you want a smaller cube that still behaves like a spectrum.
Wavelet Components keeps selected leading wavelet coefficients from each transformed spectrum.
The third dimension contains wavelet components, not physical wavelengths, so IDCubePro uses component indices 1 through N.
The spectral-axis control is therefore disabled in this mode.
For Reduced Spectrum, this is the number of reconstructed/resampled spectral channels.
For Wavelet Components, this is the number of retained wavelet coefficients.
Fewer output channels produce stronger dimensionality reduction.
The preview reports the nominal spectral dimensionality ratio as Original bands / Output bands or components.
For example, reducing 500 bands to 100 output channels corresponds to 5:1 spectral dimensionality reduction.
This is not necessarily the same as final file-size compression.
Spectral Axis After Reduction
Interpolated wavelength axis - distributes the reduced channels across the original wavelength range by interpolation.
Fast wavelength matching - invokes the standard IDCubePro channel-to-wavelength workflow.
Accurate wavelength matching - invokes the more detailed cropping-based matching workflow.
Band index - labels output channels 1 through N without assigning physical wavelengths.
Interpolated Wavelength Axis
Interpolated wavelength assignment preserves the original wavelength range but is a mathematical interpolation, not a new physical wavelength calibration.
If the input cube has no valid physical wavelength axis, IDCubePro falls back to band indices.
Before compression, missing or infinite spectral values are repaired from valid neighboring spectral samples when possible.
If a pixel spectrum contains no finite values, compression stops and reports an error.
The output cube replaces the current myData.Images working cube.
WaveletCompression metadata stores the wavelet, decomposition level, output mode, original and output band counts, compression ratio, spectral-axis handling, and original wavelengths when available.
IDCubePro refreshes the Modern GUI after compression.
With at least three output channels, separated channels are used for the RGB display. With fewer than three channels, the display switches to Single Band mode.
Wavelet compression is a dimensionality-reduction operation.
Wavelet Components are transform coefficients and must not be interpreted as wavelengths.
Reduced Spectrum remains spectral in form, but it has been wavelet-smoothed and resampled.
Strong compression can remove narrow or low-amplitude spectral features.
The progress dialog can be canceled. IDCubePro restores the original working dataset when cancellation occurs.
Use Reset Preprocessing to restore myDataOriginal when it is available.
Wavelet Compression does not overwrite the source file automatically.
Use the normal IDCubePro Save / Export workflow to save the compressed cube.
Results depend on wavelet family, decomposition level, output dimension, and preprocessing.
A high compression ratio does not by itself demonstrate that scientifically important information has been preserved.
Close Wavelet Compression help.