Band Selection

IDCubePro 2026 - LASSO / Elastic Net Band Selection

Overview

This tool identifies informative wavelengths using regularized regression.

Compares spectra from two user-selected image regions.

1. Select ROI Contrast Mode.

3. Draw ROI #1 and double-click to finish.

4. Draw ROI #2 and double-click to finish.

5. LASSO identifies wavelengths that best separate the two regions.

Automatically divides the image into spatial patches and identifies wavelengths that distinguish the patches.

Alpha = 1.0 uses LASSO and produces sparse wavelength selection.

Alpha < 1.0 uses Elastic Net and can retain groups of correlated wavelengths.

Lambda controls regularization strength.

Auto uses cross-validation to determine lambda.

Controls the number of cross-validation folds used when Lambda is set to Auto.

Non-zero coefficients indicate selected wavelengths.

Positive coefficients favor ROI #1 / class 1.

Negative coefficients favor ROI #2 / opposite class.

Creates and saves a reduced hyperspectral cube containing only the selected wavelengths.

Exports wavelength, coefficient, and selection information to a CSV file.

Removes the current training regions without deleting the LASSO results.

Clears ROIs, results, selected wavelengths, and restores the default settings.

LASSO / Elastic Net Band Selection Studio