Douglas-Peucker method is a famous method for curve generalization in automated cartography. However, this method has several disadvantages, for example it is possible that the maximal offset value of lower hierarchy is bigger than the maximal offset value of the upper hierarchy. Addressing these disadvantages, this paper suggests and realizes corresponding algorithms to overcome those shortcomings. These optimization algorithms lead to constructing multi-way tree structure which can serve as a better approach to solve generalization problems. On the basis of the multi-way structure a composite generalization method is realized.