For a project, I'm trying to count objects on an image. For the most part, I'm successful as objects are at least a little bit separated from each others.
For other portions, no morphological transformations seems to work
I'm now thinking of counting 'missing objects' : ie
source image
Idea : as all object are roughly the same size, maybe I could use an 'image template' (or some kind of np kernel)
by calculating the size of 'holes' I could calculate the num of templates that could fit ?
Any thoughts or ideas to spare ?