FFT Spectral Filtering

Overview

FFT spectral filtering operates along the spectral dimension of the hyperspectral cube using a Fast Fourier Transform (FFT).

The spectrum of each pixel is transformed into the spectral-frequency domain, selected frequency components are retained or suppressed, and the filtered spectrum is reconstructed.

FFT spectral filtering requires a valid wavelength vector for the loaded hyperspectral dataset.

The FFT Low-Pass Filter preserves slowly varying spectral components while suppressing rapidly varying components.

• High-frequency spectral noise should be reduced.
• The important spectral information varies relatively smoothly.
• Rapid band-to-band fluctuations are primarily noise.

Aggressive low-pass filtering can smooth narrow peaks, absorption features, or other rapidly changing spectral information.

The FFT High-Pass Filter emphasizes rapidly varying spectral components while suppressing broad, slowly varying background structure.

• Rapid spectral changes should be emphasized.
• Broad slowly varying background components should be suppressed.
• Narrow spectral structure is of particular interest.

High-pass filtering can amplify high-frequency noise together with the desired narrow spectral features.

The cutoff determines which spectral-frequency components are retained or suppressed. The appropriate value depends on the spectral sampling, noise characteristics, and scientific objective.

Important

FFT filtering may introduce spectral artifacts when the cutoff is too aggressive or when spectra contain abrupt discontinuities.

Workflow

1. Load a hyperspectral dataset with valid wavelength information.

2. Open Correct > Filtering & Enhancement.

3. Select FFT Low Pass or FFT High Pass.

4. Inspect the filtered spectra and image.

5. Compare the result with the original spectral features.

6. Use Reset Filters if the result is not desired.

Low-Pass: suppresses rapid spectral fluctuations.

High-Pass: suppresses broad slowly varying spectral components.