Abstract: This paper presents a new image segmentation method — FWTN (First Watershed Then Normalized cut) based on Watersheds and Graph Theory to solve the watersheds produces over-segmentation problem. FWTN uses Normalized Cut to segment between regions after use watersheds，and then produces the final segmented images．The algorithm can successfully solve over-segmentation problem，and at the same time it can improve on the NP-hard problem of Normalized Cut. Experimental results show that it is a practicable method for the image segmentation．

Key words: Image Segmentation, Watersheds, Normalized Cut, Graph Theory

摘要： 提出一种新的基于分水岭算法和图论的图像分割方法FWTN(First Watershed Then Normalized cut)，以克服传统的分水岭算法造成的图像过度分割。FWIN方法在分水岭算法之后，用Normalized Cut方法在区域之间进行分割，产生最终所分割的图像。FWIN成功地解决了过分割问题，同时还对Normalized cut方法中的NP-hard问题有所改进。实验结果表明该算法是一种切实可行的图像分割方法。

关键词: 图像分割, 分水岭, Normalized Cut, 图论