Skip to contents

Image segmentation

Source: R/utils_imagem.R
image_segment.Rd

  • image_segment() reduces a color, color near-infrared, or grayscale images to a segmented image using a given color channel (red, green blue) or even color indexes (See image_index() for more details). The Otsu's thresholding method (Otsu, 1979) is used to automatically perform clustering-based image thresholding.

  • image_segment_iter() Provides an iterative image segmentation, returning the proportions of segmented pixels.

Usage

image_segment(
  img,
  index = NULL,
  r = 1,
  g = 2,
  b = 3,
  re = 4,
  nir = 5,
  threshold = c("Otsu", "adaptive"),
  k = 0.1,
  windowsize = NULL,
  col_background = NULL,
  na_background = FALSE,
  has_white_bg = FALSE,
  fill_hull = FALSE,
  erode = FALSE,
  dilate = FALSE,
  opening = FALSE,
  closing = FALSE,
  filter = FALSE,
  invert = FALSE,
  plot = TRUE,
  nrow = NULL,
  ncol = NULL,
  parallel = FALSE,
  workers = NULL,
  verbose = TRUE
)

image_segment_iter(
  img,
  nseg = 2,
  index = NULL,
  invert = NULL,
  threshold = NULL,
  k = 0.1,
  windowsize = NULL,
  has_white_bg = FALSE,
  plot = TRUE,
  verbose = TRUE,
  nrow = NULL,
  ncol = NULL,
  parallel = FALSE,
  workers = NULL,
  ...
)

Arguments

img

An image object or a list of image objects.

index

  • For image_segment(), a character value (or a vector of characters) specifying the target mode for conversion to binary image. See the available indexes with pliman_indexes(). See image_index() for more details.

  • For image_segment_iter() a character or a vector of characters with the same length of nseg. It can be either an available index (described above) or any operation involving the RGB values (e.g., "B/R+G").

r, g, b, re, nir

The red, green, blue, red-edge, and near-infrared bands of the image, respectively. Defaults to 1, 2, 3, 4, and 5, respectively. If a multispectral image is provided (5 bands), check the order of bands, which are frequently presented in the 'BGR' format.

threshold

The theshold method to be used.

  • By default (threshold = "Otsu"), a threshold value based on Otsu's method is used to reduce the grayscale image to a binary image. If a numeric value is informed, this value will be used as a threshold.

  • If threshold = "adaptive", adaptive thresholding (Shafait et al. 2008) is used, and will depend on the k and windowsize arguments.

  • If any non-numeric value different than "Otsu" and "adaptive" is used, an iterative section will allow you to choose the threshold based on a raster plot showing pixel intensity of the index.

k

a numeric in the range 0-1. when k is high, local threshold values tend to be lower. when k is low, local threshold value tend to be higher.

windowsize

windowsize controls the number of local neighborhood in adaptive thresholding. By default it is set to 1/3 * minxy, where minxy is the minimum dimension of the image (in pixels).

col_background

The color of the segmented background. Defaults to NULL (white background).

na_background

Consider the background as NA? Defaults to FALSE.

has_white_bg

Logical indicating whether a white background is present. If TRUE, pixels that have R, G, and B values equals to 1 will be considered as NA. This may be useful to compute an image index for objects that have, for example, a white background. In such cases, the background will not be considered for the threshold computation.

fill_hull

Fill holes in the objects? Defaults to FALSE.

erode, dilate, opening, closing, filter

Morphological operations (brush size)

  • dilate puts the mask over every background pixel, and sets it to foreground if any of the pixels covered by the mask is from the foreground.

  • erode puts the mask over every foreground pixel, and sets it to background if any of the pixels covered by the mask is from the background.

  • opening performs an erosion followed by a dilation. This helps to remove small objects while preserving the shape and size of larger objects.

  • closing performs a dilatation followed by an erosion. This helps to fill small holes while preserving the shape and size of larger objects.

  • filter performs median filtering in the binary image. Provide a positive integer > 1 to indicate the size of the median filtering. Higher values are more efficient to remove noise in the background but can dramatically impact the perimeter of objects, mainly for irregular perimeters such as leaves with serrated edges.

Hierarchically, the operations are performed as opening > closing > filter. The value declared in each argument will define the brush size.

invert

Inverts the binary image, if desired. For image_segmentation_iter() use a vector with the same length of nseg.

plot

Show image after processing?

nrow, ncol

The number of rows or columns in the plot grid. Defaults to NULL, i.e., a square grid is produced.

parallel

Processes the images asynchronously (in parallel) in separate R sessions running in the background on the same machine. It may speed up the processing time when image is a list. The number of sections is set up to 70% of available cores.

workers

A positive numeric scalar or a function specifying the maximum number of parallel processes that can be active at the same time.

verbose

If TRUE (default) a summary is shown in the console.

nseg

The number of iterative segmentation steps to be performed.

...

Additional arguments passed on to image_segment().

Value

  • image_segment() returns list containing n objects where n is the number of indexes used. Each objects contains:

    • image an image with the RGB bands (layers) for the segmented object.

    • mask A mask with logical values of 0 and 1 for the segmented image.

  • image_segment_iter() returns a list with (1) a data frame with the proportion of pixels in the segmented images and (2) the segmented images.

References

Nobuyuki Otsu, "A threshold selection method from gray-level histograms". IEEE Trans. Sys., Man., Cyber. 9 (1): 62-66. 1979. doi:10.1109/TSMC.1979.4310076

Author

Tiago Olivoto tiagoolivoto@gmail.com

Examples

if (interactive() && requireNamespace("EBImage")) {
library(pliman)
img <- image_pliman("soybean_touch.jpg", plot = TRUE)
image_segment(img, index = c("R, G, B"))
}