Abstract
Segmentation of CT volume data is important and useful in non-destructive testing and evaluating. To eliminate the artifacts influence, we propose a new approach of 3D defect segmentation using two steps. First of all, an initial segmentation using 3