June 10, 2004

Contents

Watershed Segmentation

(Was available also in Scion Image Beta 3b. Not available in eariler versions, and Scion Image is no longer available.)

Watershed segmentation is a way of automatically separating or cutting apart particles that touch. It starts with a mask or binary image, where the particles are (say) black. It calculates a distance map to find the fattest parts of the object (the peaks or local maxima of the distance map). Starting with the peaks as maximal erosion points (MEP's), it dialates them as far as possible - either until the edge of the object is reached, or the edge of the region of another (growing) MEP.

A feeling for the algorithm is best obtained by watching NIH Image or ImageJ do it on a binary image made for the purpose.


This section illustrates how to make a sample image on which to try the watershed segmentation.  The drawing tools used here can be added to ImageJ.  In any event, you may want to use the image already provided:  Sample Images / Miscl / watershed_msk1.tiff

(shown half size)

(shown half size)


 

Image

 

 

Segmentation works best for smooth convex objects that don't overlap too much. The small nick in the edge caused the rectangle to be divided in half. Some of the fingers in the irregularly shaped object at the lower left do not have a narrow enough waist to cause 'cutting'.