Apparatus and method for point cloud processing
A method, apparatus or system for processing point cloud information can involve a learned deep entropy model over octrees for lossless compression/decompression of 3D point cloud data, wherein self-supervised compression/decompression involves an adaptive entropy coder operating on a tree-structured conditional entropy model and utilizing information from the local neighborhood as well as the global topology from the tree structure.
1 . An apparatus comprising a processor configured to receive a bitstream including a compressed point cloud represented as a tree structure; and to decode the bitstream by, starting from a root node of the tree structure as a current node:
constructing a context for the current node comprising at least one of a level, an octant, a location, or a parent of the current node;
predict an occupancy symbol distribution that indicates child nodes occupancy probabilities for the current node by using a learning-based entropy model based on the context for the current node and feature information from one or more available neighboring nodes and from one or more ancestor nodes of the current node; and
decode an occupancy symbol for the current node using an adaptive entropy decoder based on the occupancy symbol distribution;
expanding the tree structure based on the decoded occupancy symbol; and
output the expanded tree structure representative of a reconstruction of the compressed point cloud.
2 . The apparatus of claim 1 , wherein the tree structure is an octree, a kd-tree, a quad tree-binary tree, or a prediction tree.
3 . The apparatus of claim 1 , wherein the processor is configured to predict occupancy symbol distributions for each node of the tree structure in parallel.
4 . The apparatus of claim 1 , wherein the learning-based entropy model uses one or more deep features of one or more sibling nodes of the one or more ancestor nodes through the same learning-based model used for siblings of the current node, to predict the occupancy symbol distributions.
5 . The apparatus of claim 1 , wherein the learning-based entropy model uses one or more deep features of all nodes at one or more ancestor levels through the same learning-based model used for siblings of the current node, to predict the occupancy symbol distributions.
6 . A method comprising:
receiving an encoded bitstream including a compressed point cloud represented as a tree structure; and
decoding the encoded bitstream by, starting from a root node of the tree structure as a current node:
constructing a context for the current node comprising at least one of a level, an octant, a location, or a parent of the current node;
predicting an occupancy symbol distribution that indicates child nodes occupancy probabilities for the current node by using a learning-based entropy model and based on the context for the current node and feature information from one or more available neighboring nodes and from one or more ancestor nodes of the current node;
decoding an occupancy symbol for the current node using an adaptive entropy decoder based on the occupancy symbol distribution;
expanding the tree structure based on the decoded occupancy symbol; and
outputting the expanded tree structure representative of a reconstruction of the compressed point cloud.
7 . The method of claim 6 , wherein predicting the occupancy symbol distributions for each node of the tree structure is performed in parallel.
8 . The method of claim 6 , wherein the learning-based model uses one or more deep features of all nodes at a parent level k associated with the current node for all layers deeper than k+1.
9 . The method of claim 6 , wherein the tree structure is an octree, a kd-tree, a quad tree-binary tree, or a prediction tree.
10 . The method of claim 6 , wherein the learning-based entropy model uses one or more deep features of one or more sibling nodes of the one or more ancestor nodes through the same learning-based model used for siblings of the current node, to predict the occupancy symbol distributions.
11 . The method of claim 6 , wherein the learning-based entropy model uses one or more deep features of all nodes at one or more ancestor levels through the same learning-based model used for siblings of the current node, to predict the occupancy symbol distributions.
12 . An apparatus for compressing a point cloud represented as a tree structure; the apparatus comprising a processor configured to construct a context for each node of the tree structure comprising at least one of a level, an octant, a location, or a parent of the current node, by, starting from a root node of the tree structure as a current node:
predicting an occupancy symbol distribution that indicates child nodes occupancy probabilities for the current node by using a learning based entropy model based on the context for the current node and feature information from one or more neighboring nodes and from one or more ancestor nodes of the current node;
encoding an occupancy symbol for the current node into one of a plurality of bitstreams representative of the current node by using an adaptive entropy encoder based on the predicted occupancy symbol distribution; and
combine the plurality of bitstreams.
13 . The apparatus of claim 12 , wherein the tree structure is an octree, a kd-tree, a quad tree-binary tree, or a prediction tree.
14 . The apparatus of claim 12 , wherein the processor is configured to predict occupancy symbol distributions for each node of the tree structure in parallel.
15 . The apparatus of claim 12 , wherein the learning-based entropy model uses one or more deep features of one or more sibling nodes of the one or more ancestor nodes through the same learning-based model used for siblings of the current node, to predict the occupancy symbol distributions.
16 . A method for compressing a point cloud represented as a tree structure, the method comprising:
constructing a context for each node of the tree structure comprising at least one of a level, an octant, a location, or a parent of the current node, by, starting from a root node of the tree structure as a current node:
predicting an occupancy symbol distribution that indicates child nodes occupancy probabilities for the current node by using a learning based entropy model based on the context for the current node and feature information from one or more neighboring nodes and from one or more ancestor nodes of the current node;
encoding an occupancy symbol for the current node into one of a plurality of bitstreams representative of the current node by using an adaptive entropy encoder based on the predicted occupancy symbol distribution; and
combining the plurality of bitstreams.
17 . The method of claim 16 , wherein predicting the occupancy symbol distributions for each node of the tree structure is performed in parallel.
18 . The method of claim 16 , wherein the learning-based model uses one or more deep features of all nodes at a parent level k associated with the current node for all layers deeper than k+1.
19 . The method of claim 16 , wherein the tree structure is an octree, a kd-tree, a quad tree-binary tree, or a prediction tree.
20 . The method of claim 16 , wherein the learning-based entropy model uses one or more deep features of one or more sibling nodes of the one or more ancestor nodes through the same learning-based model used for siblings of the current node, to predict the occupancy symbol distributions.