Abstract: It is difficult for existed polygon medial axis (MA) extracting methods to address complex polygon and to make sure of the accuracy and connectivity of MA. The objective of this study is to introduce a new curved-polygon MA extracting method that is based on the mean distance transformation (MDT) of the nearest border point set (NBPS). This paper follows four steps. Firstly, on the base of simple polygon, concepts of curved-polygon together with its MA are given. Secondly, MDT of NBPS，that will be used to adjust distance values for eliminating the impact of noisy points is put forward through extending raster distance transformation (DT) method. Thirdly, efficient judgment regulations which are constructed according to elements of DT and characteristics of MA points, and the method of greed points-growth and detection are used to complete the mission of extracting MA. Furthermore, specific steps and processes to achieve extracting MA are given. Fourthly, various complex polygons are used to inspect this new method. It is found that the impacts of border noisy points are eliminated effectively and both the accuracy and connectivity of MA are satisfactory which overcome the shortcomings of traditional methods.