White Blood Cell (WBC) nucleus segmentation is a fundamental step in automated hematological image analysis, as the accuracy of subsequent feature extraction and classification largely depends on precise nucleus localization. However, variations in staining protocols, illumination conditions, and image acquisition systems often reduce the robustness of conventional grayscale-based segmentation methods. This paper presents a robust and fully unsupervised framework for WBC nucleus segmentation based on Intuitionistic Fuzzy Divergence (IFD) thresholding in the Lab color space. Unlike the conventional IFD approach, which performs thresholding on grayscale images, the proposed method applies the IFD optimization process to the a-channel of the Lab color space, where leukocyte nuclei exhibit significantly higher chromatic contrast than surrounding blood components. The optimal threshold is determined by minimizing the IFD, and the extracted binary mask is subsequently used to delineate nucleus regions. The proposed method was evaluated on three independent peripheral blood smear image datasets acquired under different staining protocols and imaging conditions. Experimental results demonstrate that the proposed framework maintains consistent segmentation performance across heterogeneous datasets. Visual comparisons further indicate that the proposed approach produces more reliable nucleus segmentation than conventional grayscale-based IFD thresholding. Owing to its unsupervised nature, low computational complexity, and robustness to imaging variations, the proposed framework provides an effective preprocessing solution for computer-aided leukocyte analysis and automated hematological diagnosis.