Currently, the generalization of map compilation still primarily depends on the manual operation, except the inconsistency caused by the operators, the procedure is also time-consuming. As for the digital generalization, there is some available software. However, because of the different complexities of features, we still could not have effective control on the parameters. By this research, we try to classify the different types of coastline and to figure out their detailed characteristics. Then by adopting the MGE Map Generalizer (MGMG) developed by Intergraph Co., with the algorithms such as Douglas & Peucker, the Nth Point Thinning, Lang, Reumann-Witkam, Point Relaxaton and VectGen to test the generalization of coastline. The series topographic maps are adopted as the standard maps, and the generalization is performed from the scale 1:50000 with different parameters, then the generalized maps are compared with those of smaller scale to decide suitable parameters. From the test and comparison, we tried to analyze and induce a set of generalization parameters which is suitable for the coastline of Taiwan area. The outcomes could not only solve the quality inconsistency in the conventional map compilation, but also shorten the time for map production.