An evolutionary-based entropic image thresholding approach for nano-scale light microscopic image segmentation

Abstract

In this paper an Evolutionary-based hybrid thresholding method is presented and implemented on nano-scale light microscopic images. Because of background non-uniform illumination, Segmentation of nano-scale light microscopic images is a hard task in real world, and also fundamental task in image processing. An adaptive and efficient thresholding method based on image spatial correlation histogram and Shanbag entropy is proposed in this paper. Genetic algorithm as a parameter optimizing tool is also employed to fine-tune the parameters and coefficients. The microscopic nano-scale images of rat prostate cancer cells with the spatial resolution of few tens of nanometers and nuclear track images (few tens to few hundred nanometers in spatial resolution) are segmented by the proposed thresholding method and the misclassification error and track detection rate are used as the criteria for evaluation purpose. The results exhibit the efficiency and capability of the proposed method in thresholding the real world image dataset.