Automated blob detection using iterative Laplacian of Gaussian filtering and unilateral second-order Gaussian kernels

Detecting overlapping blob objects is a classical, yet challenging problem in the image processing field. In this paper, we propose an automated blob detection method that is able to tackle both isolated and partially overlapping blob objects. Firstly, we present a multiscale normalization method for Laplacian of Gaussian kernels, thus proposing iterative Laplacian of Gaussian filtering to attenuate the overlapping regions of the adjacent blobs while retaining the isolated blobs. Secondly, we investigate the potential of unilateral second-order Gaussian kernels for separating overlapping blobs, and explain how to set the scales of the kernels appropriately. Eventually, the blob detection result can be easily obtained by a thresholding procedure. We have applied the proposed method to fluorescence microscopy cell images and electron micrography nanoparticle images. The experimental results demonstrate that the proposed method outperforms the competing methods including state-of-the-art methods for dealing with partially overlapping blob objects. (C) 2019 Elsevier Inc. All rights reserved.