sobal: Sobel-Feldman operator

In jeffreyevans/spatialEco: Spatial Analysis and Modelling Utilities

Sobel-Feldman operator

Description

An isotropic image gradient operator using a 3x3 window

Usage

sobal(x, method = "intensity", ...)

Arguments

x	A raster class object
method	Type of operator ("intensity", "direction", "edge")
...	Additional arguments passed to raster::overlay or, if method="edge", raster::focal (if you want a file written to disk use filename = "" argument)

Details

The Sobel-Feldmanh operator is a discrete differentiation operator, deriving an approximation of the gradient of the intensity function. abrupt discontinuity in the gradient function represents edges, making this a common approach for edge detection. The Sobel-Feldman operator is based on convolving the image with a small, separable, and integer matrix in the horizontal and vertical directions. The operator uses two 3x3 kernels which are convolved with the original image to calculate approximations of the derivatives - one for horizontal changes, and one for vertical. Where x is defined here as increasing in the right-direction, and y as increasing in the down-direction. At each pixel in the raster, the resulting gradient can be combined to give the gradient intensity, using: SQRT( Gx^2 Gy^2 ). This can be expanded into the gradient direction using atan(Gx/Gy)

Value

A raster class object or raster written to disk

Author(s)

Jeffrey S. Evans <[e-mail]>

References

Sobel, I., & G. Feldman, (1969) A 3x3 Isotropic Gradient Operator for Image Processing, presented at the Stanford Artificial Intelligence Project (SAIL).

Examples

library(terra)

r <- rast(system.file("ex/logo.tif", package="terra"))  
  s.int <- sobal(r[[1]])
  s.dir <- sobal(r[[1]], method = "direction")
  s.edge <- sobal(r[[1]], method = "edge")

opar <- par(no.readonly=TRUE)
par(mfrow=c(2,2))
  plot(r[[1]])
  plot(s.int, main="intensity") 
  plot(s.dir, main="direction") 
  plot(s.edge, main="edge")
par(opar)   

jeffreyevans/spatialEco documentation built on April 4, 2024, 10:53 a.m.