Comparative analysis on automated separation methods of clustered regions composed of convex particles

Automated splitting of overlapping convex particles,especially the complex clumps,remains an important and challenging issue in image analysis,which directly affect the holistic performance.We compared and analyzed the primary steps,advantages and disadvantages,and applicability of three kinds of clump splitting methods,i.e.the concavity analysis,ellipse fitting and watershed transform.The results of anlysis indicate that(1) the splitting algorithm based on concavity analysis is limited to the simple cluster composed of few and convex particles;(2) the ellipse fitting separation method is apt for automatic count of few and convex particles in simple clusters;(3) the watershed transform is a powerful and fast algorithm to split clumps,which has the advantages of being parallelizable,always producing a complete division,and especially efficient and effective for clumps consisted of a large number of components with symmetric or regular shapes.This is beneficial for the design of image analysis or pattern recognition system related to cluster splitting.