Image Analysis completing shapes in a binary image

Posted by Obsn on Jun 26, 2015; 9:30am
URL: http://imagej.273.s1.nabble.com/Image-Analysis-completing-shapes-in-a-binary-image-tp5013313.html

Hi guys,

I am analyzing the growth of ice crystals and I want to create an automated image analysis procedure. I process the images as follows:
- converting to 8-bit grayscale
- setting the threshold to create a binary image
- filling holes in the binary image (the ice crystals are bounded by a dark edge and their inside is as bright as the background of the image)
- watershed-split (splitting clusters)
- Measuring and counting the crystals

My problem is that some crystals don’t show a homogeneous dark edge. In some spots the edge is too bright for the set threshold values. But I can’t change the threshold values because the background noise gets too high. So the results in the binary images for these inhomogeneous crystals are C-shaped crystals instead of the needed O-shaped crystals. It is necessary that the crystals are O-shaped, so that they can be filled and split.
Is there any way to complete these C-shaped crystals, like a reversed watershed-split? Dilating, a defined number of cycles, until the edges touch and then erode until the original size of the crystals are reached and creating a line in between the edges which didn’t touch. Or is there any other way to complete these crystals in order to get the expected shape?
Thanks in advance and I’d really appreciate your help.

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html