Point Cloud Attribute Encoding With RAHT Parent Node Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current point cloud compression methods, such as VPCC and GPCC, face inefficiencies in encoding and decoding attributes, particularly in handling large datasets and dynamic point clouds, leading to suboptimal compression ratios and increased data storage needs.
Innovation Solution
The proposed method employs a voxel-based structure with Region-Adaptive Hierarchal Transform (RAHT) to determine true and predicted transform coefficients, calculating residuals and encoding them for improved compression efficiency by selectively using parent nodes based on attribute differences, thereby enhancing prediction accuracy and reducing residuals in the bitstream.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional image/video MPEG encoders are used in VPCC approach, then existing coders can be reused and dynamic point clouds are supported, but compression gain is limited and the method is hardly usable for static point clouds
Solution Approach 1:
The patent segments the point cloud processing into two independent parts: geometry encoding using octree structure and attribute encoding using RAHT. This segmentation allows each part to use optimized algorithms suitable for its specific requirements, improving overall compression efficiency while maintaining support for both static and dynamic point clouds
Solution Approach 2:
The patent introduces adaptive prediction mechanisms that dynamically adjust to the characteristics of the point cloud data. The prediction process adapts to local variations in attribute values, and the encoding strategy can be adjusted based on whether the point cloud is static or dynamic, optimizing compression for each case
2Ease of operation
If RAHT is applied to all parent nodes without selection, then the encoding process is simple, but prediction accuracy decreases in heterogeneous regions leading to larger residuals
Solution Approach 1:
The patent applies local quality by differentiating the encoding approach based on local characteristics of the point cloud data. In homogeneous regions, simple encoding is used, while in heterogeneous regions, adaptive prediction with parent node selection is employed. This ensures optimal compression efficiency tailored to local data characteristics without uniformly complicating the entire encoding process
Solution Approach 2:
The patent changes the parameter of prediction accuracy by selectively choosing parent nodes based on attribute similarity. Instead of using a fixed simple prediction method, the system dynamically adjusts the prediction accuracy parameter by selecting from multiple parent nodes depending on the local heterogeneity, thereby optimizing the balance between complexity and accuracy
3Productivity
If all parent nodes are used for prediction, then the encoding is computationally efficient, but prediction accuracy deteriorates in regions with significant attribute variations
Solution Approach 1:
The patent performs preliminary action by pre-evaluating the suitability of parent nodes for prediction before actually performing the encoding. The system assesses attribute similarities and selects appropriate parent nodes in advance, which allows it to achieve high prediction accuracy without having to evaluate all possible parent nodes during the main encoding process, thus maintaining computational efficiency
Data Source
Figure 1
Figure 2
Figure 3a~3d
AI summary
Method for encoding and decoding attributes of a point cloud, as well as encoder and decoder. Therein, the method for encoding attributes of points of a point cloud to generate a bitstream of compressed point cloud data, wherein the point cloud's geometry is represented by a voxel-based structure with a plurality of nodes having parent-child relationships by recursively splitting a volumetric space containing the point cloud into sub-volumes each associated with a node of the voxel-based structure, comprises the steps of: Determining a true transform coefficient by applying Region-Adaptive Hierarchal Transform, RAHT, to the attributes of a present node; Determining a difference between the attributes of a present parent node and each node of a first set of nodes sharing a face or an edge with the present node, wherein the present parent node is the parent node of the present node; Selecting a second set from the first set of parent nodes according to the difference, wherein preferably the second set of parent nodes includes the present parent node; Determining a predicted value for the attributes of the present node according to the attributes of the second set of parent nodes; Determining a predicted transform coefficient by applying RAHT to the predicted value of the attributes of the present node; Determining a residual from the true transform coefficient and the predicted transform coefficient; and Encoding the residual to produce encoded data of attributes of the point cloud for the bitstream.