Thresholding

[Rotella, Anthony M. (GRC)[]]

Hello,

        I am currently working on some research at NASA doing Image
Analysis on CT-Scanned Solder Joints, looking into voiding defects.
Basically, the images are two circles, one dark one in the center (the
resistor lead) and one outer circle(grayish, the solder), with small
voids everywhere. These joints were created in microgravity so air
bubbles, etc..don't escape the joint due to the lack of buoyant forces.
Anyway, I wrote an algorithm that manages to pretty decently change the
resistor lead to the background color, and separate the pores from the
new background. The problem is that some voids touch the surface of the
void, so when the image is thresholded, they just merge with the
background and then don't show up as an individual object when I use the
"Analyze Particle" feature. Given this, I thought it may be possible to
write an algorithm to create a "tertiary" image, making the voids grey,
keeping the background black, and making the solder white. Then it would
be easy to count up all the 127's, getting the void area. Unfortunately,
it's messier than I thought, since the transition from solder to black
background doesn't go smoothly straight to 0 intensity. I was wondering
if there were any suggestions on how to write such a thresholding
algorithm, or a way to get around this transition from background to
foreground so I don't get this silly ring of gray around the joint. When
the algorithms ImageJ uses are run, it doesn't get this ring or
anything, I am also curious to hear if it doesn't anything besides just
changing the pixel values within the thresholding range. Thanks a lot,

        Tony Rotella