Point Cloud Coding Using Hierarchical Group of Frames
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing point cloud coding techniques face limitations in prediction accuracy and coding efficiency due to the use of a single reference frame for inter prediction, which restricts the utilization of redundant information across frames and results in suboptimal coding performance, especially when transmission resources are limited.
Innovation Solution
The proposed method involves dividing frames into groups of frames (GOFs) and using multiple reference PC samples within the same GOF for inter prediction, allowing for both earlier and later timestamped frames to improve prediction accuracy and coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single reference frame is used for inter prediction, then device complexity is reduced, but prediction accuracy and coding efficiency deteriorate
Solution Approach 1:
The patent divides the reference frame structure into multiple segments by introducing Groups of Frames (GOFs) with different types (Type 0, Type 1, Type 2). Each GOF type can use different reference frame configurations, allowing the system to segment the complexity management while maintaining high prediction accuracy where needed. This resolves the contradiction by organizing multiple reference frames into structured groups rather than using a single undifferentiated reference frame.
Solution Approach 2:
The patent introduces dynamic switching between different GOF types based on coding conditions and requirements. The encoder can dynamically select which reference frames to use within each GOF type, and can switch between GOF types to adapt to different scene characteristics. This dynamic approach allows the system to maintain low complexity when possible while achieving high prediction accuracy when needed, resolving the static contradiction between simplicity and performance.
2Measurement precision
If multiple reference frames are used for inter prediction, then prediction accuracy and coding efficiency are improved, but device complexity increases
Solution Approach 1:
The patent applies different reference frame strategies to different local regions or different GOF types based on their specific requirements. Type 0 GOFs use one reference frame configuration while Type 1 and Type 2 GOFs use different configurations. This local differentiation allows the system to use multiple reference frames and achieve high prediction accuracy only where beneficial, rather than uniformly across all frames, thus managing overall complexity while improving accuracy where needed.
Solution Approach 2:
The patent changes key parameters of the reference frame structure by introducing GOF-type-specific parameters such as reference frame indices, prediction modes, and hierarchical relationships. By parameterizing the reference frame usage according to GOF types, the system can efficiently manage multiple reference frames through structured parameter control rather than ad-hoc management, reducing the complexity overhead of using multiple reference frames while maintaining their predictive benefits.
3Productivity
If frames with later timestamps are used as reference frames, then coding performance is improved, but causality and decoding order challenges arise
Solution Approach 1:
The patent prepares reference frames in advance by organizing them into GOF structures where later-timestamp frames are pre-encoded and stored as reference material. Type 2 GOFs specifically allow using later-timestamp frames as references by pre-processing them. This preliminary preparation allows the current frame to benefit from high-quality prediction using future frames without causing decoding timing issues, as the future frames are already available in the reference buffer from previous encoding passes or alternative decoding paths.
Data Source
AI summary
Embodiments of the present disclosure provide a solution for point cloud coding. A method for point cloud coding is proposed. The method comprises: determining, during a conversion between a current PC sample of a point cloud sequence and a bitstream of the point cloud sequence, one or multiple reference PC samples for the current PC sample, wherein at least one reference frame comprising the one or multiple reference PC samples and a current frame comprising the current PC sample are in a GOF; and performing the conversion based on the one or multiple reference PC samples.