Predictive Geometry Coding in G-PCC for Low-Loss Point Clouds

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing point cloud compression techniques, such as G-PCC, face challenges with computationally intensive context-coded bins due to lossy coding in the (r, ϕ, i) domain and require efficient methods to reduce rounding errors.

Innovation Solution

Implement predictive geometry coding in the (r, ϕ, i) domain with a second set of residuals in the Cartesian domain to minimize loss and computational intensity, using angular mode for LIDAR-acquired point clouds.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If coding is performed in the (r, φ, i) domain with angular mode, then compression efficiency is improved for LIDAR-acquired point clouds, but rounding errors are introduced causing lossy compression

Engineering Contradiction:
Improvecompression efficiencyVSAvoidposition accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent segments the residual coding into two parts: primary residuals coded in the (r, φ, i) domain for compression efficiency, and secondary residuals coded in the Cartesian domain for precision correction. This segmentation allows each part to serve its specific function optimally.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent combines two different coding approaches (angular mode coding and Cartesian residual coding) into a composite solution. The primary residuals use angular mode for efficiency while secondary residuals use Cartesian coordinates for accuracy, creating a hybrid approach that leverages the strengths of both methods.

Inventive Principle:
Principle #40Composite materials

2Measurement precision

If a second set of residuals in Cartesian domain is added to reduce rounding errors, then position accuracy is improved, but computational complexity increases

Engineering Contradiction:
Improveposition accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies local quality by using different coding strategies for different parts of the data. Primary residuals use the more efficient angular mode coding, while only the necessary secondary residuals in Cartesian domain are coded to correct specific rounding errors. This selective approach maintains accuracy without unnecessarily increasing complexity.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent changes the parameter domain for different residual components. Primary residuals are coded in angular parameters (r, φ, i) while secondary residuals are coded in Cartesian parameters (x, y, z). This parameter change allows the system to optimize for both compression and accuracy as needed.

Inventive Principle:
Principle #35Parameter changes

3Adaptability or versatility

If context coded bins are used to signal primary and secondary residuals, then coding flexibility is improved, but computational intensity increases significantly

Engineering Contradiction:
Improvecoding flexibilityVSAvoidcomputational intensity
Core Design Contradiction:
Adaptability or versatilityVSPower

Solution Approach 1:

The patent extracts the essential accuracy-critical information into secondary residuals and codes them separately using context-coded bins. By taking out only the necessary correction terms and coding them efficiently, the system maintains flexibility while reducing the overall computational burden compared to coding all residual information with full context models.

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP4226628B1Predictive geometry coding in g-pcc
Publication Date: 2025.08.20 QUALCOMM INC
  • EP4226628B1 patent drawingFigure 1
  • EP4226628B1 patent drawingFigure 2
  • EP4226628B1 patent drawingFigure 3

AI summary

An example method of decoding a point cloud includes selecting, from a plurality of predefined prediction modes, a prediction mode for performing predictive geometry coding of a position of a current node of the point cloud, wherein the plurality of prediction modes includes at least: a zero prediction mode, and a delta prediction mode; responsive to selecting the zero prediction mode: determining a radius, an azimuth, and a laser index of a parent node of the current node; inferring an azimuth and a laser index of a predicted position of the current node as the azimuth and the laser index of the parent node; inferring a radius of the predicted position to be a minimum radius value, wherein the minimum radius value is different than the radius of the parent node; and determining, based on the predicted position of the current node, the position of the current node.