MNF Explorer
The MNF Explorer performs Maximum Noise Fraction analysis of hyperspectral images and provides component visualization, clustering, manual selection, and export tools.
IDCubePro 2026 - MNF Explorer
Overview
The MNF Explorer performs Maximum Noise Fraction analysis of hyperspectral images and provides component visualization, clustering, manual selection, and export tools.
MNF transforms the original spectral data into components ordered according to signal relative to estimated noise.
The MNF Composite displays three selected MNF components as an RGB image.
Red Component - MNF component assigned to the red display channel.
Green Component - MNF component assigned to the green display channel.
Blue Component - MNF component assigned to the blue display channel.
Use the sliders or numeric fields to change the MNF components displayed in each color channel.
The weight fields control the relative contribution of the red, green, and blue MNF components to the composite image.
A weight of 1 preserves the default contribution.
Higher values increase the contribution of that channel.
Contrast adjusts the display range of each MNF component.
Increasing the value suppresses extreme low and high values and can improve visualization of intermediate structures.
Contrast changes the display only and does not modify the MNF calculation.
Gamma changes the brightness response of the displayed MNF composite.
Gamma affects visualization only and does not modify the MNF scores.
Restores the default MNF composite settings: MNF1 / MNF2 / MNF3, channel weights = 1, contrast = 0, and gamma = 1.
The Component Spectra panel displays the spectral loading vectors associated with the MNF components currently assigned to the red, green, and blue channels.
The horizontal axis represents wavelength when wavelength information is available; otherwise it represents band or channel number.
The MNF Component Quality graph displays the generalized MNF eigenvalues.
Higher eigenvalues indicate components with stronger signal relative to the estimated noise covariance.
Components are ordered from highest to lowest MNF eigenvalue.
The reference line at eigenvalue = 1 is provided as a visual aid for evaluating component quality.
The clustering section groups pixels according to their coordinates in MNF component space.
Current clustering uses MNF1, MNF2, and MNF3.
Select the desired number of K-medoids clusters.
The current interface supports 2 to 25 clusters.
Select the distance metric used by K-medoids clustering.
Squared Euclidean is a useful general starting point.
Click Run Clustering to perform K-medoids clustering of the MNF score data.
The resulting classes are displayed in both the Cluster Image and Cluster Scatter panels.
The Cluster Image shows the spatial distribution of the K-medoids classes in the original image coordinates.
Each cluster is displayed using a different color.
The Cluster Scatter panel displays the MNF score distribution used for clustering.
2D displays MNF1 versus MNF2.
3D displays MNF1, MNF2, and MNF3.
Changing between 2D and 3D redraws the existing clustering result and does not rerun K-medoids.
After clustering, use the class dropdown to select an individual cluster.
Selecting Class 1, Class 2, and so on creates a spatial mask for that class and displays the selected pixels in the Cluster Image.
Selecting All Classes restores the complete clustering result.
Lasso provides manual selection in MNF score space.
3. The scatter plot switches to 2D MNF1 versus MNF2.
4. Draw a polygon around the desired group of MNF points.
The selected MNF points are mapped back to their original image pixels and displayed in the Cluster Image.
The Lasso selection becomes the active selection for export.
Clear removes the current class or Lasso selection and restores the complete clustering display.
The existing K-medoids clustering result is preserved, so clustering does not need to be rerun.
Exports the original hyperspectral datacube with pixels outside the active class or Lasso selection set to zero.
The original spectral dimensions and associated metadata are preserved.
Exports the complementary region.
Pixels inside the active selection are excluded, while pixels outside the selection are preserved.
Exports the active spatial selection as a binary mask.
The exported mask can be used for segmentation, quantitative analysis, machine learning labels, or later processing.
Exports the complete MNF score datacube.
Each image plane corresponds to one MNF component.
1. Inspect the default MNF1 / MNF2 / MNF3 composite.
2. Examine the MNF Component Quality graph.
3. Explore additional MNF components using the sliders.
4. Adjust component weights, contrast, or gamma if needed.
5. Choose the number of clusters and distance metric.
7. Inspect the clustering result in 2D and 3D.
8. Select an individual class or use Lasso for manual selection.
9. Export the selected region, inverse region, or binary mask.
10. Export the complete MNF cube if further MNF analysis is required.