G-PCC Occupancy Coding via Inter Prediction and Planar Mask
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing point cloud encoding and decoding technologies face challenges in efficiently compressing and decompressing 3D point cloud data, particularly in handling inter prediction and planar mode data, which affects the encoding of single occupancy values.
Innovation Solution
The Geometry Point Cloud Compression (G-PCC) encoder employs inter-prediction by calculating a global motion vector to generate a prediction frame and encodes local motion vectors to reposition points accurately. It also uses planar mode to determine contexts for encoding occupancy data, avoiding the encoding of single occupancy values when inter prediction and/or planar mode data are encoded.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If single occupancy values are encoded for all nodes, then occupancy information is complete, but bitstream size increases and processing complexity increases
Solution Approach 1:
The patent extracts and removes redundant occupancy information from the bitstream. When a node is marked as occupied and its children are decoded, the occupancy values of child nodes are inferred and don't need to be explicitly encoded. This extraction of redundant data reduces bitstream size while maintaining complete occupancy information through inference mechanisms.
Solution Approach 2:
The patent performs preliminary marking of occupied nodes before decoding child nodes. By预先 marking which nodes are occupied, the decoder can infer the occupancy status of child nodes without requiring explicit encoding, thereby reducing the amount of data that needs to be transmitted and processed.
2Measurement precision
If inter prediction and planar mode data are encoded, then prediction accuracy is improved, but processing requirements increase
Solution Approach 1:
The patent applies different coding strategies to different regions of the point cloud based on local characteristics. Inter prediction and planar mode are selectively applied to nodes where they provide significant accuracy improvement, rather than uniformly to all nodes. This localized application reduces overall processing requirements while maintaining high prediction accuracy where needed.
Solution Approach 2:
The patent applies prediction techniques partially - only to occupied nodes and their descendants, rather than to all nodes in the octree. This partial application reduces processing requirements while still achieving accurate prediction for the relevant portions of the point cloud data.
3Measurement precision
If all occupancy values are explicitly coded, then decoding accuracy is ensured, but bitstream size increases
Solution Approach 1:
The patent extracts redundant occupancy information from the bitstream by inferring child node occupancy status from parent node occupancy markers. This extraction eliminates the need to explicitly encode occupancy values for all nodes, significantly reducing bitstream size while ensuring decoding accuracy through the inference mechanism.
Solution Approach 2:
The patent uses occupancy information from parent nodes to copy or infer the occupancy status of child nodes. Instead of transmitting separate occupancy values for each node, the system copies the occupancy concept from parent to child through the prediction framework, reducing the quantity of data transmitted while maintaining decoding accuracy.
Data Source
AI summary
An example device for coding point cloud data includes a memory configured to store point cloud data; and one or more processors implemented in circuitry and configured to: determine whether inter prediction data is coded for a current node of an octree of the point cloud data; determine whether planar mask data is coded for the current node; when at least one of the inter prediction data or the planar mask data is coded for the current node, avoid coding a single occupancy value for the current node, the single occupancy value indicating whether only a single sub-node of the current node includes a point; and code the current node. The processors may also be configured to determine a context for entropy coding the planar mask data according to planar mask data for a collocated node in a reference frame when the planar mask data is coded.