A hybrid active contour model using local region-based K-medoids for infrared image segmentation

Pengqiang Ge, Minjie Wan, Yunkai Xu, Xiaofang Kong, Yixuan Kang, Guirong Weng, Guohua Gu, Qian Chen

Abstract: Highlights•The multi-feature bias fields are pre-estimated through LRKM module before LSE.•Intensity, texture, roughness and gradient features are used to depict IR targets.•The weights between local and global data-driven terms are automatically assigned.•The energy amplitude of the hybrid term is regularized to boost system robustness.