Point Cloud Spherical Quantization for Uniform Cartesian Sampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing point cloud codecs, particularly G-PCC, suffer from sub-optimal compression efficiency due to uniform quantization of azimuthal angles and radius in spherical coordinates, leading to non-uniform sampling and reconstruction errors in cartesian space, especially for sparse geometry data captured by spinning Lidar.
Innovation Solution
Adaptive quantization of azimuthal angles based on radius, using non-uniform quantization steps Δϕ(r2D) to achieve uniform sampling in cartesian space, combined with uniform quantization of circular arcs Δϕarc, optimizing compression performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If uniform quantization is applied to azimuthal angles and radius in spherical coordinates, then the encoding process is simple, but the compression efficiency is sub-optimal due to non-uniform sampling in cartesian space
Solution Approach 1:
The patent changes the quantization parameter from uniform to adaptive based on radius. Specifically, the quantization step for azimuthal angle is modified from a constant value to a variable value that depends on the radius r2D, using the formula Δϕ(r2D) = Δϕarc / r2D. This parameter change ensures uniform sampling density in cartesian space while maintaining encoding simplicity.
Solution Approach 2:
The patent applies different quantization strategies to different regions of the point cloud based on their distance from the reference point. Points at different radii receive different quantization steps, with closer points getting finer quantization and farther points getting coarser quantization. This local adaptation optimizes compression efficiency for each region independently.
2Ease of operation
If uniform quantization steps are used for azimuthal angles, then the quantization process is straightforward, but reconstruction errors occur in cartesian space due to non-uniform sampling
Solution Approach 1:
The quantization step parameter is changed from a constant to a radius-dependent variable. The relationship Δϕ(r2D) = Δϕarc / r2D ensures that the arc length represented by each quantization step remains constant (Δϕarc), which maintains uniform sampling density in cartesian space and reduces reconstruction errors.
Solution Approach 2:
The patent transitions from uniform quantization in spherical coordinate space to uniform quantization in cartesian space by introducing a radius-dependent transformation. This dimensional transformation ensures that the quantization is uniform with respect to the actual spatial distance rather than the angular coordinate alone.
3Productivity
If adaptive quantization based on radius is applied to azimuthal angles, then compression efficiency improves, but the quantization complexity increases
Solution Approach 1:
The patent introduces a single adaptive parameter (the radius-dependent quantization step) that simplifies the overall system by providing a clear mathematical relationship between radius and quantization precision. This parameter change achieves compression efficiency improvement without requiring complex adaptive algorithms.
Solution Approach 2:
The radius values are computed beforehand during the spherical coordinate transformation, and these pre-computed radius values are then used to determine the appropriate quantization steps. This preliminary computation avoids the need for complex real-time adaptive quantization decisions during the encoding process.
Data Source
AI summary
There is provided methods and apparatus of encoding/decoding a point cloud into/from a bitstream of encoded point cloud data representing a physical object, each point of the point cloud being associated with spherical coordinates representing an azimuthal angle responsive to a capture angle of a sensor that captured the point and a radius responsive of a distance of the point from a referential. The encoding method includes adaptively quantizing a residual error azimuthal angle, derived from the azimuthal angle, based on the radius.