Variation of the watershed algorithm for summit detection

The summitWatershed function implements a segmentation strategy to identify summits within a landscape image generated by the PathwaySpace package. This function is entirely coded in R, which helps alleviating users from the task of loading an excessive number of dependencies. Nonetheless, while this novel implementation prevents the burden a 'dependency heaviness', it still requires optimization as it currently exhibits slower performance compared to well-established implementations such as the watershed function from the EBImage package. The summitWatershed maintain a certain level of compatibility with the EBImage's watershed function, and both can be used in the PathwaySpace package.

Usage

summitWatershed(x, tolerance = 0.1, ext = 1)

Arguments

x

A 2D-numeric array in which each point represents the coordinates of a signal in a landscape image.

tolerance

The minimum signal intensity of a summit (in [0,1]), representing a fraction of the maximum signal intensity.

ext

Radius (in pixels) for detecting neighboring objects.

Value

A matrix with labeled summits.

See also

summitMapping

Author

Sysbiolab Team, Mauro Castro.

Examples

# Load a demo landscape image
data('gimage', package = 'PathwaySpace')

# Scale down the image for a quicker demonstration
gimage <- gimage[200:300, 200:300]

# Check signal range
range(gimage, na.rm = TRUE)
#> [1] 0.2733407 0.9760202
# [1] 0 1

# Check image
# \donttest{
image(gimage)

# }

# Threshold the signal intensity, for example:
gimage[gimage < 0.5] <- 0

# Run summit segmentation
gmask <- summitWatershed(x = gimage)

# Check resulting image mask
# \donttest{
image(gimage)

# }