Point cloud compression methods based on deep learning are being studied. Extracting a part of point cloud from compressed data to grasp the characteristic shape of an object is an important function. I propose ranking and hierarchical coding that can sequentially decode main features of point cloud.