Accurate segmentation of White Blood Cells (WBCs) is considered one of the most critical steps in automated hematological disease diagnosis systems and blood cell classification. However, variations caused by imaging conditions, staining techniques, illumination intensity, and the presence of overlapping cells significantly degrade the accuracy and generalizability of many existing methods. In this study, a robust approach for WBC segmentation in microscopic blood smear images is proposed based on three invariant characteristics of peripheral blood images. In the proposed method, the background mask is initially extracted using the M channel of the CMYK color space combined with the Zack thresholding algorithm. Subsequently, WBC nuclei are identified, and Red Blood Cell (RBC) regions are determined by simultaneously exploiting background and nuclear information. Then, preliminary WBC regions are extracted using the color characteristics of RBCs. To separate clustered WBCs, a modified Marker-Controlled Watershed algorithm based on distance transform and a refined nuclear mask is employed to prevent over-segmentation. Finally, WBC-like artifacts are eliminated using geometric features. The performance of the proposed method is evaluated on three standard datasets with different imaging conditions and staining protocols. The results demonstrate that the proposed method can provide accurate and stable WBC segmentation independent of the dataset type, while exhibiting strong robustness against variations in imaging conditions and the presence of overlapping cells.