Touching Objects Count with Progressive Erosion

Abstract

Counting touching objects is of great importance to accurately count all objects in images. Morphology erosion is a simple, effective and efficient operation, commonly used in binary image analysis. The progressive erosion with changing image can accurately count touching objects by separating touching objects into isolated ones. This work also supports that the distance transform for counting touching objects can be replaced by a certain progressive erosion. With the help of counting strategies, the proposed method, which counts touching objects based on concave points, can be applied to object counting in many situations. Experimental results indicate that the proposed algorithm is both effective and efficient.