Skip to contents

Estimate Image Haze for Dark Object Subtraction (DOS)

Source: R/estimateHaze.R
estimateHaze.Rd

estimates the digital number (DN) pixel value of *dark* objects for the visible wavelength range.

Usage

estimateHaze(
  x,
  hazeBands,
  darkProp = 0.01,
  maxSlope = TRUE,
  plot = FALSE,
  returnTables = FALSE
)

Arguments

x

SpatRaster or a previous result from estimateHaze with returnTables = TRUE from which to estimate haze

hazeBands

Integer or Character. Band number or bandname from which to estimate atmospheric haze (optional if x contains only one layer)

darkProp

Numeric. Proportion of pixels estimated to be dark.

maxSlope

Logical. Use darkProp only as an upper boundary and search for the DN of maximum slope in the histogram below this value.

plot

Logical. Option to display histograms and haze values

returnTables

Logical. Option to return the frequency table per layer. Only takes effect if x is a SpatRaster. If x is a result of estimateHaze tables will always be returned.

Value

If returnTables is FALSE (default). Then a vector of length(hazeBands) containing the estimated haze DNs will be returned. If returnTables is TRUE a list with two components will be returned. The list element 'SHV' contains the haze values, while 'table' contains another list with the sampled frequency tables. The latter can be re-used to try different darkProp thresholds without having to sample the raster again.

Details

It is assumed that any radiation originating from *dark* pixels is due to atmospheric haze and not the reflectance of the surface itself (the surface is dark, i.e. it has a reflectance close to zero). Hence, the haze values are estimates of path radiance, which can be subtracted in a dark object subtraction (DOS) procedure (see radCor)

Atmospheric haze affects almost exclusively the visible wavelength range. Therefore, typically, you'd only want to estimate haze in blue, green and red bands, occasionally also in the nir band.

Examples

## Estimate haze for blue, green and red band
haze <- estimateHaze(lsat, hazeBands = 1:3, plot = FALSE)
haze
#> B1_dn B2_dn B3_dn 
#>    55    19    12 

## Find threshold interactively
#### Return the frequency tables for re-use
#### avoids having to sample the Raster again and again
haze <- estimateHaze(lsat, hazeBands = 1:3, returnTables = TRUE)
## Use frequency table instead of lsat and fiddle with
haze <- estimateHaze(haze, hazeBands = 1:3, darkProp = .1, plot = FALSE)
haze$SHV
#> B1_dn B2_dn B3_dn 
#>    57    20    12