I’m attaching a pipeline which I think will get your closer to what you want. Your pipeline was definitely pointed in the right correction, so I made the following modifications:
- Some pre-processing to get rid of the background illumination heterogeneity.
- With your images, trying to identify the full droplet boundary didn’t seem to work well, due to their touching in some cases, but mostly due to the bright “hole” in the center of each droplet. But it was easier to identify the “hole” first,and then expand out to the full droplet after.
- Using filtering to distinguish the “holes” from everything else, much as you already did.
Also, you will want to remove the scale bar from the images prior to running the pipeline.
2013_02_19.cp (11.4 KB)