Neighbor-Based Point Cloud Coding With Geometry Model Conversion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Point cloud data, due to its large size, poses challenges for efficient storage and transmission, necessitating effective compression techniques that maintain quality and accuracy.
Innovation Solution
A method for encoding point cloud data using a tree-based or triangle-based geometry representation model, where geometry information of neighboring sub-volumes is converted and predictive coding is employed to enhance compression efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If point cloud data is compressed using conventional methods, then data size is reduced, but prediction accuracy and compression efficiency deteriorate due to insufficient utilization of neighboring geometry information
Solution Approach 1:
The point cloud volume is divided into multiple sub-volumes, and each sub-volume is further divided into octants. This segmentation allows independent processing and prediction of each region using its own geometry representation model, enabling efficient utilization of local geometric information while maintaining overall compression efficiency.
Solution Approach 2:
Geometry representation models of neighboring sub-volumes are converted to the same type as the current sub-volume before prediction. This preliminary conversion ensures that the prediction process can directly utilize neighboring geometric information without type mismatch issues, improving prediction accuracy and compression efficiency.
2Measurement precision
If geometry representation models of neighboring sub-volumes are converted to match the current sub-volume type, then prediction accuracy improves, but processing complexity increases
Solution Approach 1:
The geometry representation models are converted by changing their type parameters to match the current sub-volume's model type. This parameter change approach allows flexible adaptation between different geometry representation types (e.g., from one octant-based model to another) while maintaining a unified processing framework, thus improving prediction accuracy without excessive complexity increase.
3Productivity
If more neighboring geometry information is utilized for predictive coding, then compression ratio improves, but coding complexity increases
Solution Approach 1:
The coding process segments the point cloud into sub-volumes and octants, allowing prediction to be performed independently in each segment using its own geometry representation model. This segmentation enables efficient utilization of neighboring geometric information across multiple scales without overwhelming complexity, as each segment processes only its local information.
Solution Approach 2:
The geometry representation models of neighboring sub-volumes are pre-converted to match the current sub-volume type before the actual prediction and coding steps. This preliminary action prepares the data in advance, allowing the coding process to directly use the converted models without additional type-checking or conversion overhead during compression, thus improving compression ratio while controlling complexity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Systems, apparatuses, methods, and computer-readable media are described herein for determining and/or coding geometry information. Point cloud information (e.g., of a point cloud associated with content) may be predicted. A first plurality of sub-volumes of a point cloud may, for example, be coded based on a second plurality of sub-volumes of a reference point cloud. Geometry information of the second plurality of sub-volumes (e.g., represented by a second type of geometry model) may be used to code geometry information of the first plurality of sub-volumes (e.g., represented by a first type of geometry model).