Wavelet Compression

Features:

This function performs compression in combination with spectral cropping .The Wavelet Compression function performs wavelet-based compression on hyperspectral data and generates a new dataset. It prompts the user to enter the type of wavelet and the level of decomposition. The hyperspectral data is loaded, and the wavelet transform is applied to each pixel's spectral signature. The resulting compressed data is saved in a separate file.

Step 1.

When the file is opened, select Wavelet compression method: Edit → Compress → Wavelet compression.

The function will prompt you to enter the type of wavelet (e.g., 'db1') and the level of decomposition (a positive integer) from the dropdown menues. Specify the wavelet type and decomposition level in the input dialog and click Apply. The function will apply the wavelet transform to the hyperspectral data, compressing it.

Currently supported wavelets families: Haar, Daubechies wavelets, Symlets, Coiflets

Step 2.

After setting parameters, a popup dialogue will ask for a folder to store the wavelet processed file. After compression, a dialog box will appear showing the wavelet type, level of decomposition, the size of the saved file.

The message box indicates that the new compressed datacube file can be opened in a usual way. Click OK on the dialog box. Another dialog box will appear indicating that compression is complete. The name of the file will be generated automatically as avelet_compressed_datacube_db1_3.mat, where the db1 corresponds to the type of wavelet and “3” to the level of decomposition.

Step 3.

Open the saved file and explore the spectrum of the region of interest. You will find that the spectrum seems to be “compressed”.

NOTE: after a wavelet transform, the bands in the spectral analysis window do not correspond to the actual physical values such as wavelengths.

Step 4.

Perform Spectral Cropping. Using Data Tips from the Strip Toolbar identify the end of the spectrum as shown below and enter a similar value in the Upper Limit field in the Spectral Crop panel. Click Apply. You can save the datacube as a new dataset. The new dataset will have significantly lower number of bands and therefore significantly smaller size. 

NOTE: different types of wavelets and different level of decomposition will provide a different level and quality of compression.