Image Threshold Segmentation Based on GLLE Histogram

Chunming He, Xiaobo Wang, Lizhen Deng, Guoxia Xu

Abstract: To improve the threshold segmentation performance, we propose a novel 2-D histogram with the measure of gray level and local entropy. The proposed GLLE histogram introduces local entropy, which can effectively describe the spatial information of pixels. Furthermore, we put forward the criterion function for GLLE histogram and calculate the threshold by Renyi entropy, which makes our method more generalized. Experiment results show that our method is effective and robust compared with seven other methods in segmentation accuracy and has the ability to resist the illumination.