Partially occluded object detection and counting

Abstract

In today's fast-moving life many a pathological labs still use probabilistic approach to count the number of microscopic cells in a sample slide, which is a tedious job and not accurate all the time. The problem of finding the position of objects in an image as well as to count the occurrence of that object in the image is a hitch for the researcher. We have restricted our problem only to completely circular disk shaped objects. Our aim is to accurately calculate the geometrical position of the objects and the object count. Boundaries of the object patches have been extracted using morphological operation and with the help of classical geometry we locate the exact boundary coordinate of the objects. In this paper we propose a new, simple framework for visual object detection and object counting for the image of circular objects, whether these are partially occluded to each other or not.