When calibrating multi-camera systems with existing camera calibration toolboxes, it is oftentimes required that calibration boards are fully captured by all cameras to establish correspondence of X-junctions automatically. It becomes impractical when different cameras share limited common field of views, resulting in the incomplete samples of the calibration board. This article proposes a practical solution containing a modified checkerboard pattern and image processing algorithms. Corresponding world and image points for calibration algorithms are provided by the proposed solution when calibration boards are partially captured. X-junctions are categorized in two types and recognized in a group as an identification unit. The proposed pattern is arranged to ensure uniqueness of each identification unit; thus X-junctions can be positioned by supporting algorithms and correspondences are founded. No burden is introduced in sampling compared to traditional calibration methods. Experimental results of calibrating a multi-camera system verify that the use of the partially captured images brings benefits to calibration accuracy. Furthermore, the proposed method can substantially facilitate the calibration process.