Point Cloud Decoding Using Hierarchical Occupancy Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing point cloud encoding and decoding technologies, such as those using the octree method, fail to improve compression performance for sparse point clouds measured by LiDAR, despite achieving improvements for dense point clouds.
Innovation Solution
A point cloud decoding device and method that predicts occupancy information of child nodes using the occupancy information of child hierarchical nodes, incorporating scanning range and interval data from LiDAR sensors to enhance intra prediction accuracy and compression performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If occupancy information of 26 parent hierarchical nodes is referred to for intra prediction, then compression performance is improved for dense point clouds, but processing time increases and no effect is achieved for sparse point clouds
Solution Approach 1:
The patent applies local quality by differentiating the prediction approach based on point cloud density. For sparse regions, it uses a simplified method referring to fewer child hierarchical nodes, while for dense regions, it can utilize more nodes. This localized adaptation optimizes processing time for sparse point clouds while maintaining compression performance where needed.
Solution Approach 2:
The patent segments the prediction process into different hierarchical levels. Instead of uniformly referring to 26 parent hierarchical nodes for all child nodes, it performs prediction at the child hierarchical node level using a limited set of occupied child hierarchical nodes, thereby reducing the overall processing burden while maintaining effective compression.
2Loss of information
If occupancy information of 26 parent hierarchical nodes is referred to for intra prediction, then compression performance is improved for dense point clouds, but device complexity increases
Solution Approach 1:
The patent simplifies the prediction process by applying local quality - using a focused, localized approach that refers to only occupied child hierarchical nodes rather than all 26 parent hierarchical nodes. This reduces device complexity while maintaining compression effectiveness for the specific characteristics of sparse point clouds.
3Loss of time
If the number of parent hierarchical nodes referred to is reduced to seven, then processing time is reduced, but compression performance deteriorates for dense point clouds
Solution Approach 1:
The patent introduces dynamics by adaptively selecting the number of child hierarchical nodes to refer to based on the actual occupancy pattern. Rather than fixing the number at seven, it dynamically adjusts by referring to occupied child hierarchical nodes, which can be fewer than seven in sparse regions (reducing processing time) while potentially utilizing more nodes in dense regions (maintaining compression performance).
Solution Approach 2:
The patent changes the parameter of node reference from a fixed count (seven) to a dynamic count based on occupancy status. This parameter change allows the system to optimize between processing time and compression performance by adapting to the actual data characteristics rather than using a static configuration.
Data Source
AI summary
A point cloud decoding device (200) according to the present invention including a circuit, wherein the circuit: stores occupancy information of a child hierarchical node indicating whether or not the child hierarchical node is occupied; and predicts occupancy information of a child node indicating whether or not the child node is occupied using the occupancy information of the child hierarchical node.