Compute Minimum Noise Fraction transform

Compute Minimum Noise Fraction transform

Usage

hsi_calc_mnf(x, trim = 0L, ...)

Arguments

x

A SpatRaster with hyperspectral data.

trim

Non-negative integer. Number of bands to drop symmetrically from both spectral edges before computing the transform. Default 0L applies no trimming.

...

Additional arguments passed to spacetime::mnf(), notably Sigma.Noise for a custom noise covariance matrix and use for NA handling. Default noise estimate uses 0.5 * cov(diff(x)) (MAF).

Value

An object of class c("mnf", "prcomp") with components:

values

Numeric vector of eigenvalues (noise fractions), one per band.

rotation

Numeric matrix of eigenvectors (loadings).

x

Numeric matrix of MNF scores (pixels x bands).

Details

MNF components are ordered by decreasing noise fraction: MNF1 carries the most noise, the final components carry the most signal. Eigenvalues approximate the noise fraction under the proportional covariance model and are non-negative, but may exceed 1 with real data. 1 - eigenvalue is the lag-1 autocorrelation of that component. Inspect $values to identify the signal-rich tail before passing the result to hsi_apply_mnf().

When x has been pre-processed with hsi_smooth_savgol(), spectral edge bands become nearly collinear due to the polynomial fitting, which can make the noise covariance matrix singular. Set trim to ceiling(n / 2) where n is the window size passed to hsi_smooth_savgol() to drop the affected bands from both ends before computing the transform.

Wraps spacetime::mnf() by Edzer Pebesma, implementing the algorithm of Green et al. (1988) with noise estimation following Switzer & Green (1984).

Green, A.A., Berman, M., Switzer, P. and Craig, M.D. (1988). A transformation for ordering multispectral data in terms of image quality with implications for noise removal. IEEE Transactions on Geoscience and Remote Sensing, 26(1), 65–74.

Switzer, P. and Green, A. (1984). Min/max autocorrelation factors for multivariate spatial imagery. Technical Report, Department of Statistics, Stanford University.

See also

hsi_apply_mnf(), spacetime::mnf()

Other HSI Transformations: hsi_apply_mnf(), hsi_calc_difference(), hsi_calc_ndi(), hsi_calc_raba(), hsi_calc_rabd(), hsi_calc_ratio(), hsi_calc_rcv(), hsi_calc_reflectance(), hsi_calc_remp(), hsi_calc_rmean(), hsi_calc_rmedian(), hsi_calc_rsd(), hsi_calc_stretch(), hsi_destripe(), hsi_remove_continuum(), hsi_smooth_median(), hsi_smooth_savgol(), hsi_tiled(), hsi_write_scaled()

Examples

if (FALSE) { # \dontrun{
x <- terra::rast("REFLECTANCE_testdata.tif")
x_mnf <- hsi_calc_mnf(x)
x_mnf$values

# After hsi_smooth_savgol() with window = 15
x_mnf <- hsi_calc_mnf(x, trim = 8L)
} # }