This paper suggested a new efficient lossless compression algorithm for vector maps in view of the graphical characteristics. First, we adopt the differential lossless transformation for the graphic data in the vector map files, to make the original data which should have been represented by double model or float model be represented by long model. Then, based on the abbreviated compression, the prediction coding based on the irregular coefficients is used. Namely, it is to dynamically distribute proper byte storage to the transformed coefficients which have been differentiated according to the numerical values, and compressed further combined with a dictionary coding method. Finally, we do an experimental contrast for the compression effect.