Point Cloud Geometry Encoding With Contextual Vertex Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing point cloud compression technologies, such as V-PCC and G-PCC, struggle to efficiently encode and decode the geometry of point clouds, especially for dense and sparse point clouds, leading to inefficient data transmission and storage, particularly in applications like AR/VR where high resolution and real-time transmission are required.
Innovation Solution
The method and device utilize contextual information to select a coding probability for entropy coding, using occupancy and positional information of neighboring cuboid volumes and edges to encode and decode vertex positions in a point cloud represented by cuboid volumes modeled with triangles, avoiding simple bypass coding of vertex positions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing point cloud compression technologies (V-PCC and G-PCC) are used to encode geometry, then data can be compressed, but compression efficiency is insufficient for dense and sparse point clouds
Solution Approach 1:
The patent performs preliminary classification of vertices into boundary vertices (on cuboid surfaces) and internal vertices (inside cuboid volumes) before encoding. This preliminary action allows different encoding strategies to be applied to different vertex types, improving overall compression efficiency by treating vertices according to their spatial characteristics rather than using a uniform approach
Solution Approach 2:
The patent segments the point cloud geometry encoding into distinct components: boundary vertex position encoding, internal vertex position encoding, and attribute encoding. This segmentation allows each component to be optimized independently, with boundary vertices encoded using contextual information from neighboring cuboid volumes and internal vertices encoded using different strategies, thereby resolving the contradiction between compression efficiency and data fidelity
2Measurement precision
If high resolution point cloud data is transmitted for AR/VR applications, then quality is improved, but transmission efficiency and real-time capability deteriorate
Solution Approach 1:
The patent changes the parameter representation by encoding vertex positions relative to cuboid volume boundaries rather than using absolute coordinates. Boundary vertex positions are encoded as offsets from cuboid surfaces, and internal vertex positions are encoded using contextual probabilities derived from neighboring cuboid occupancy information. This parameter transformation enables more efficient compression at high resolutions, maintaining quality while improving transmission efficiency for real-time AR/VR applications
3Device complexity
If simple bypass coding is used for vertex positions, then encoding complexity is reduced, but compression efficiency deteriorates
Solution Approach 1:
The patent applies local quality by using different encoding methods for different spatial locations. Boundary vertices on cuboid surfaces use contextual probability models based on neighboring cuboid occupancy information, while internal vertices use different probability models. This localized approach to encoding complexity allows the system to achieve better compression efficiency without uniformly increasing device complexity across all vertex types
Data Source
AI summary
A method of encoding geometrical information of a geometry of a point cloud into a bitstream is provided. The point cloud is represented by a plurality of cuboid volumes, an occupied cuboid volume being modelled by one or more triangles. The method includes: for a vertex on a current edge: constructing contextual information based on at least one of: occupancy information of neighboring cuboid volumes that abut the current edge, or vertex positional information of already-coded neighboring edges of the current edge, the neighboring edges being edges having a point in common with the current edge; using the contextual information to select a coding probability of an entropy coder; and encoding, by the entropy coder and using the selected coding probability, a vertex position of the vertex on the current edge.