Hierarchical automated clustering of cloud point set by ellipsoidal skeleton: application to organ geometric modeling from CT-scan images

摘要

Abstract: We present a robust method for automaticallyconstructing an ellipsoidal skeleton (e-skeleton) froma set of 3D points taken from NMR or TDM images. Toensure steadiness and accuracy, all points of theobjects are taken into account, including the innerones, which is different from the existing techniques.This skeleton will be essentially useful for objectcharacterization, for comparisons between variousmeasurements and as a basis for deformable models. Italso provides good initial guess for surfacereconstruction algorithms. On output of the entireprocess, we obtain an analytical description of thechosen entity, semantically zoomable (local featuresonly or reconstructed surfaces), with any level ofdetail (LOD) by discretization step control in voxel orpolygon format. This capability allows us to handleobjects at interactive frame rates once the e-skeletonis computed. Each e-skeleton is stored as a multiscaleCSG implicit tree. !35