Adaptive QtBt Decomposition for Point Cloud Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current graph-based point cloud compression (G-PCC) methods are inefficient for point clouds with almost symmetric bounding boxes, as they enforce asymmetric decomposition, which can be counterproductive and limit flexibility in partitioning, leading to compromised compression efficiency.
Innovation Solution
The proposed method employs a quad-tree-binary-tree (QtBt) decomposition that allows for flexible partitioning by switching among Octree, Quadtree, and Binary-tree at each level, with parameters K and M to determine the decomposition type, enabling efficient compression even for symmetric bounding boxes by skipping unnecessary occupancy information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If asymmetric decomposition is enforced in current G-PCC methods, then decomposition can be performed for all point clouds, but compression efficiency is compromised for point clouds with symmetric bounding boxes
Solution Approach 1:
The patent applies dynamics by making the decomposition method adaptive rather than fixed. The system dynamically selects between asymmetric decomposition, symmetric decomposition, and no decomposition based on the actual characteristics of the bounding box (aspect ratios of x, y, z dimensions). This allows the decomposition approach to change according to the input data properties, resolving the contradiction between enforcing a specific method and adapting to different point cloud types.
Solution Approach 2:
The patent changes the parameter of decomposition symmetry based on bounding box characteristics. By evaluating aspect ratios and comparing them against thresholds, the system adjusts the decomposition strategy (asymmetric vs. symmetric) to match the geometric properties of the point cloud data. This parameter adaptation enables optimal compression efficiency for both symmetric and asymmetric bounding boxes.
2Productivity
If octree decomposition is used for symmetric bounding boxes, then decomposition is simple and uniform, but unnecessary occupancy information is encoded reducing compression efficiency
Solution Approach 1:
The patent extracts only the necessary occupancy information for encoding by determining the appropriate decomposition type first. For symmetric bounding boxes, it extracts minimal occupancy data by using symmetric decomposition or skipping decomposition entirely, rather than encoding all occupancy information that would result from forced asymmetric octree decomposition. This extraction principle reduces the quantity of encoded data while maintaining compression efficiency.
Solution Approach 2:
The patent applies partial action by performing decomposition only when necessary (asymmetric bounding boxes) and skipping it for symmetric cases. This partial application of decomposition avoids the excessive encoding of redundant occupancy information while still providing sufficient partitioning for compression when the geometry requires it.
3Productivity
If flexible partitioning with QtBt decomposition is implemented, then compression efficiency improves for asymmetric bounding boxes, but device complexity increases due to multiple decomposition types
Solution Approach 1:
The patent manages complexity by changing a single key parameter (decomposition type selection) based on simple geometric criteria (aspect ratios). Rather than implementing complex adaptive logic, the system uses straightforward parameter comparison against thresholds to determine whether to use asymmetric decomposition, symmetric decomposition, or no decomposition. This parameter-based approach maintains relatively simple device complexity while achieving flexible partitioning.
Solution Approach 2:
The patent applies local quality by applying different decomposition strategies to different regions of the parameter space (different bounding box aspect ratio combinations). Each local region (symmetric, asymmetric, highly asymmetric) receives the appropriate decomposition treatment, rather than applying a single complex algorithm universally. This localized approach simplifies the overall system design.
Data Source
AI summary
A method of point cloud attribute coding is performed by at least one processor and includes receiving a point cloud data set comprising a bounding box. A determination is made as to whether the bounding box is symmetric or asymmetric. A quad-tree-binary-tree (QtBt) decomposition is performed on the point cloud data set based on the bounding box being asymmetric.