A novel labelling algorithm for object counting

Counting applications use various image processing techniques to count the objects of interest. The final and the most significant step in any image-based counting application is labeling. It becomes extremely difficult to label when the objects of interest overlap. Existing counting applications use labeling algorithms that provide the same label to overlapping objects. In this paper, we propose a new labeling algorithm which has the ability to distinguish certain overlapping objects and faithfully label both disjoint and overlapping objects. This results in more precise counting results vis-à-vis the present labeling algorithms.