Plugins

This is a set of ImageJ plugins for color image segmentation. The set features four plugins:

• The original K-Means plugin from Jarek Sacha's IJ Plugin Toolkit (that you can also find here)
• A refactored version of the original K-Means plugin providing color space selection (RGB, XYZ, Lab*, HSB), a simpler initialization criterion and a few more visualization modes
• A Fuzzy C-Means variant introducing another initialization criterion plus a new visualization mode
• A Spatial version of the Fuzzy C-Means algorithm that considers, for each pixel, a window containing the neighbor pixels in order to compute the membership function

Color images can be converted into four different color spaces (considering RGB) leading to differently clustered images:

• XYZ a standardized color space, similar to RGB but with less correlated components
• Lab* a perceptional uniform color space that provides human predictable segmentation results
• HSB similar to Lab* for its purpose of giving a more human readable color representation, but differs in the representation of the luminance component and consequently of the chroma

• Cluster Centroid Color each cluster has been assigned its centroid color (in case of color conversion the color space is converted back to RGB)
• Gray Scale each pixel is labelled with the number of the cluster it belongs to, and the image range is stretched in 0-255
• Random RGB a random RGB color is generated for each cluster, it may be very useful when many clusters share a similar centroid color value
• Binary Stack a stack of binary images is outputted, each of those representing a cluster: a black pixel belongs to the current slice cluster, a white not. It may be useful to extract back the original region with an OR operation
• Fuzzy Stack a stack of gray scaled images is used to show the membership values of each pixel to each cluster, the darker the pixel in a slice, the more the corresponding pixel in the original images belongs (in a fuzzy sense) to the currently selected cluster

Just put the jar file (that you can find in the /Releases/ directory of the project) under the /plugins directory of ImageJ and relaunch it. The plugins will appear in the Plugins>Segmentation menu. A micro guide explaining the user interface can be found in the /doc/ directory.

• Antonio Vergari
• Francesco Tangari

26/12/2010 Refactored the K-Means Plugin according to the new design (now the algorithm is much more independent form ImageJ API)

30/12/2010 Created a new package from the upcoming Fuzzy C-Means version of the plugin. Started to work on it.

3/01/2011 Started to work on the Spatial version of the Fuzzy C-Means algorithm.

15/01/2011 Released version 1.0 of the plugin suite with several bugs fixed and a working spatial implementation

5/02/2011 Released version 1.1 implementing a way to disable the random seeding