Version: 0.9.1
Author: Hannah Weller

An R package that counts colors within color range(s) in images, and provides a masked version of the image with targeted pixels changed to a different color. Output includes the locations of the pixels in the images, and the proportion of the image within the target color range with optional background masking. Users can specify multiple color ranges for masking.

For more information, read the introduction on CRAN.

About countcolors

This package is the result of a collaboration between Sarah Hooper, Sybill Amelon, and me (Hannah Weller), with the goal of quantifying the area of white-nose syndrome infection of bat wings.

Countcolors allows users to quantify regions of an image by color(s). It’s a companion package to colordistance, which provides quantitative color comparisons of images. While both of these packages borrow techniques from image processing, they don’t use any machine learning, adaptive thresholding, or object detection. This makes them reliable and easy to use, but limited in application.

Countcolors is best suited for images where the color(s) of interest are reasonably different from the rest of the image – so if you’ve got an image of some treetops, it can quantify the amount of green tree cover, but it wouldn’t necessarily be able to distinguish a species with leaves of one slightly different shade of green from all the others due to variations in lighting (among other sources of variation). If this is the kind of specificity you need, you’re better off just mucking in and using real image processing. I recommend the OpenCV library for C++ or Python.

December 27, 2018: Fixed bugs in countColors that threw errors when specifying filepaths for save.indicator or providing numeric values for target.color. Thanks to Lee Yuhui for discovering them.

Installation instructions

  1. Install the devtools package (install.packages("devtools")).
  2. Install countcolors using the install_github function:

Questions or feedback?

Email me: