In this paper, we propose 2D segmentation algorithm for 3D visualization using CT bone images. Kmeans algorithm and thresholding method are traditional methods that can detect human bone. But the noise appeared on segmented images generated by both Kmeans algorithm and thresholding method. To detect bone cluster, the two assumptions are considered, they are as follows; first, pixels that have high intensity can be merged into bone cluster. Secondly, because a bone cluster consists of large or small group, the ambiguous pixels can be merged into bone cluster by estimating segmentation function. With the proposed FSTB(Flexible Segmentation To Boundary) algorithm the noise didn't appeared on segmented CT bone images and the algorithm carried out effective segmentation to the ambiguous pixels.