Projector calibration is crucial for the structured light three-dimensional measurement system. However, the projector is unable to capture images like a camera, thus the accuracy of projector calibration is determined by the mapping relationship between the pixel coordinate system of a camera and that of a projector. The goal of optimizing projector calibration is to find a more accurate mapping relationship. We propose a sub-pixel mapping method based on blob detection to find the corresponding projector pixel coordinates of the marker points of a calibration target and then realize sub-pixel level projector calibration. This method can reduce eccentric error and is also suitable for the projectors with different pixel arrangements. Besides, it is independent of the results of camera calibration. Experimental results demonstrate that our proposed mapping method can achieve good projector calibration performance: its re-projection error is about 10% less than the existing methods and the accuracy of measurement with the calibrated results is also higher.