Computing refined skeletal features from medial point clouds

摘要: We present methods for computing refined features from 3D medial-surface point clouds.Features include: medial classification, surface decomposition into sheets, Y-network extraction, and robust regularization.We compute our efficiently robustly using as input only a unstructured cloud.We illustrate the use of in shape classification patch-based part-based segmentation. Medial representations have been widely used many analysis processing tasks. Large complex shapes are, this context, challenging case. Recently, several proposed that extract point-based surfaces with high accuracy computational scalability. However, resulting clouds are limited due to difficulty such clouds. In paper, we show how bridge gap between having raw cloud enriching feature points, medial-point axis regularization, extraction. further properties can be support sample applications including edge detection segmentation, wide range shapes.