分割虚拟人切片数据的SVM多类分割方法

SVM multi-class segmentation method for slice data of virtual human

针对虚拟人切片数据量大、解剖结构复杂等特点,对分割虚拟人切片图像的基于二叉树SVM多类分割方法进行研究。基于二叉树的SVM多类分割方法较其他SVM多分类方法更符合人们分割虚拟人切片图像的习惯,而且能获得较高的分割性能和质量。通过对该方法的性能分析,为组织高效的二叉树SVM多类分割方法提供了理论支持。

This paper performed a study of SVM mutli-class segmentation method based on binary tree to segment the huge, complex slice data of virtual human. The mentioned multi-class method could get a better performance and result. It is more suitable for the human behavior to segment the slices. The analysis of performance gave a theoretical base to construct a more effective multi-classifier based on binary tree for segmentation the slice images of the virtual human.