一种基于模糊连接度和维诺图的混合分割方法

A HYBRID SEGMENTATION METHOD USING FUZZY CONNECTEDNESS AND VORONOI DIAGRAM CLASSIFICATION

介绍一种可用于医学图像处理的、集成了模糊连接度和维诺图分类算法的混合分割方法。首先采用模糊连接度算法对指定图像区域进行过滤处理形成组织样本数据,这些输出数据将作为维诺图分类算法的输入数据和分类标准,然后通过维诺图分类算法对其进行迭代处理直至形成近似的图像区域边界。最终的输出值为一组分割后的三维图像数据,可以采用体绘制方法形成三维图像分割结果,也可用于进一步的图像处理。和其他医学图像分割方法相比,这种混合分割方法集成了基于区域和基于边界两种不同的分割方法,兼具两者的优点,通过两种分割方法的协同工作,提高了图像分割的精度,适用于复杂图像的分割处理。在医学图像计算机辅助诊断系统中集成了这一方法并取得了良好的实际应用效果。

This paper presents a hybrid segmentation method which integrates fuzzy connectedness and Voronoi diagram classification algorithms and is suitable for medical images processing.We start with the fuzzy connectedness filter to perform filtering treatment on designated region of images and to generate the data of tissue sample.These output data are to be used as the input data and the classification standard for Voronoi diagram classification algorithm,and then they are iterated with Voronoi diagram classification algorithm until the approximate boundary of image region is resulted.Final output is a set of segmented 3D image data that can be used to display the 3D result of the segmentation image using volume rendering techniques,or be passed to another filter for further image processing.Compared with other medical images segmentation methods,this hybrid segmentation method integrates region-based and boundary-based segmentation methods and has the advantages of the both.The synergy of two different segmentation methods tends to result in higher segmentation quality and suits some complicated segmentation tasks.We have already integrated this approach in our medical images computer-aided diagnosis system and got a satisfying result in actual application.