In pre-processing, when transforming hyperspectral data to a Minimum Noise Function(MNF) data, the dimensionality will reduce and at the same time the wavelength information will loose. B'z MNF is one dimensionality reduction technique and will give the best bands only. Then how we will compare the MNF spectral curve with USGS spectral library using ENVI software. Any Tutorial video available?
In essence, the process is quite straightforward. Typically, we employ MNF to eliminate noise and reduce dimensions. Subsequently, we determine the number of viable and beneficial bands post this procedure. However, to effectively utilize these chosen bands and compare them with the spectral library, the following solution needs to be executed (these steps pertain to Envi 5.6):
Inverse MNF Rotation > Select Input File: The MNF file which has been changed in dimension (or you can select those bands in the Spectral Subset section) > Select the MNF statistical file with the extension ".sat" > Specify the output location
Upon completing these steps, a new file devoid of noise and reduced in dimension is generated while retaining the wavelength information of the original data.
Comparing the USGS spectral library with Minimum Noise Fraction (MNF) data in hyperspectral image analysis using ENVI software involves several steps. First, import your hyperspectral data into ENVI and apply the MNF transformation to reduce noise and enhance spectral signatures. Then, extract spectral signatures from specific regions of interest (ROIs) in your MNF-transformed image. Next, import the USGS spectral library into ENVI and match the extracted spectra with those in the library. Evaluate similarity using statistical measures or visual inspection, interpret the results to identify materials or substances in the image, and validate findings with ground truth data. Following these steps allows for effective comparison and analysis, facilitating valuable insights from hyperspectral imagery.
Jabar Habashi Thank you so much.
First i done the Forward MNF and selected the first 50 bands.
After that, tried Inverse rotation MNF. But not getting the reduced bands(50 bands) in the final output image(instead of reduced bands i'm getting the same number of bands as in the input reflectance image ie, 158). How to resolve?
Work flow
Select 'Inverse MNF Rotation' > 'Select Input File': (MNF file with reduced bands, 50 selected ) > 'Select the MNF statistical file' (selected the file with extension ".sat") > Enter the output location
Prosenjit Kayal Thank you so much
At the time of MNF transformation, there is one issue i.e., when we selecting the first noise free bands(first 50 bands), the wavelength information will loose. Then how we will compare the extracted spectral signatures from MNF-transformed image with USGS spectral library(it contains whole range of spectral information). I tried the spectral resampling techniques also, the same above mentioned issue is there.
This indicates that your data's dimensionality has been reduced. A new file has been generated with 50 effective dimensions, maintaining the same spectral range as the original data, and that's OK
Jabar Habashi , https://www.nv5geospatialsoftware.com/docs/MinimumNoiseFractionTransform.html
To find noise-free bands or reduce the band numbers, we must select the 'Number of Components' field in the 'Inverse MNF Rotation' operation. If not, all bands will be used.
This option is not visible to me in ENVI 5.6. Any ideas on this?
- Badges
- Science topic
- Similar topics
- Chemistry