Abstract: This paper presents a new spine segmentation method using Normalized cut method and OTSU method.At first，a rough segmentation result is got using normalized cut method，and an effective region is extracted according to the shape feature.Then，OTSU segmentation and edge detection method are applied in the effective region to get edges of inner components.Experimental results show that the proposed method can detect the edge of all vertebrae accurately，and eliminate the influence of artificial edges.“Segmentation on demand” is presented to guarantee unnecessary processing on background region is bypassed，which leads to more rapid processing.

Key words: Normalized cut method, OTSU method, spine image, graph-partition, segmentation on demand

摘要： 提出一种融合Normalized Cut方法和OTSU方法的脊椎图像分割算法。首先通过Normalized Cut方法得到粗略的分割结果，然后根据脊椎的形状特征提取出有效的脊椎区域，在有效的脊椎区域进行OTSU分割算法，并通过边缘检测算法得到脊椎内部的轮廓边界。实验结果证明该方法可以较为精确地检测出椎骨轮廓的边界，有效地消除了伪边缘的影响。提出了按需分割的概念，首先消除背景区域的影响，然后在有效区域进行更为精细的处理，这样可以较大程度提高算法的处理速度。

关键词: Normalized cut方法, 大津法, 脊椎图像, 图分割, 按需分割