segmentation: Image Segmentation

Description

Basic image segmentation like connected components labelling, blob extraction and fuzzy c-means

Usage

image_connect(image, connectivity = 4)
image_split(image, keep_color = TRUE)
image_fuzzycmeans(image, min_pixels = 1, smoothing = 1.5)

Arguments

image: magick image object returned by image_read() or image_graph()
connectivity: number neighbor colors which are considered part of a unique object
keep_color: if TRUE the output images retain the color of the input pixel. If FALSE all matching pixels are set black to retain only the image mask.
min_pixels: the minimum number of pixels contained in a hexahedra before it can be considered valid (expressed as a percentage)
smoothing: the smoothing threshold which eliminates noise in the second derivative of the histogram (higher values gives smoother second derivative)

Details

image_connect Connect adjacent pixels with the same pixel intensities to do blob extraction
image_split Splits the image according to pixel intensities
image_fuzzycmeans Fuzzy c-means segmentation of the histogram of color components

image_connect performs blob extraction by scanning the image, pixel-by-pixel from top-left to bottom-right where regions of adjacent pixels which share the same set of intensity values get combined.

Examples

Run this code

# Split an image by color
img <- image_quantize(logo, 4)
layers <- image_split(img)
layers

# This returns the original image
image_flatten(layers)

# From the IM website
objects <- image_convert(demo_image("objects.gif"), colorspace = "Gray")
objects

# \donttest{
# Split image in blobs of connected pixel levels
if(magick_config()$version > "6.9.0"){
objects |>
  image_connect(connectivity = 4) |>
  image_split()

# Fuzzy c-means
image_fuzzycmeans(logo)

logo |>
  image_convert(colorspace = "HCL") |>
  image_fuzzycmeans(smoothing = 5)
}
# }

Run the code above in your browser using DataLab

Get 50% off unlimited learning