Point Cloud Sequence Encoding Through 2D Regularized Plane Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing point cloud sequence encoding technologies face inefficiencies due to high spatial sparsity and destruction of spatio-temporal correlation, particularly in octree-based and video-based methods, leading to increased data redundancy and reduced encoding efficiency.
Innovation Solution
A point cloud sequence encoding method utilizing two-dimensional regularized plane projection, which includes global motion compensation, two-dimensional projection, and motion estimation to enhance spatio-temporal correlation, reducing redundancy through geometric and attribute information encoding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If octree-based encoding is used for point cloud sequences, then the encoding structure can handle sparse data, but empty nodes occupy high proportion and spatio-temporal correlation cannot be fully reflected, reducing encoding efficiency
Solution Approach 1:
The patent transforms the three-dimensional point cloud data into two-dimensional projection views (front view, side view, top view). This dimensionality change allows the encoding to capture spatio-temporal correlations more effectively in the projected 2D space, avoiding the sparsity problem in 3D octree structures while maintaining the ability to handle sparse data through multiple viewing angles
Solution Approach 2:
The patent divides the point cloud sequence into multiple frames and further segments each frame into different projection views (front, side, top). This segmentation allows independent encoding of each view while maintaining temporal correlations between frames, improving overall encoding efficiency by processing manageable segments rather than the entire 3D point cloud at once
2Productivity
If video-based point cloud encoding is used, then dense point clouds can be encoded efficiently, but cubic projection destroys spatio-temporal correlation of some areas, affecting sparse point cloud encoding efficiency
Solution Approach 1:
The patent uses orthogonal projection to transform 3D point cloud coordinates (x, y, z) into 2D projection plane coordinates, creating front view (x-y plane), side view (y-z plane), and top view (x-z plane). This dimensionality reduction preserves spatio-temporal correlations by maintaining spatial relationships in multiple 2D projections, avoiding the information loss caused by cubic projection while improving encoding efficiency
Solution Approach 2:
The patent creates multiple copies of the point cloud data in different projection views (front view, side view, top view). Each view is a 2D projection copy that captures different aspects of the 3D spatial relationships. By encoding multiple copies from different angles, the method preserves spatio-temporal correlations that would be lost in a single cubic projection, while still achieving efficient compression through standard 2D video coding techniques
Data Source
Figure 1
Figure 2
Figure 3~5
AI summary
The present invention discloses a point cloud sequence encoding/decoding method and apparatus based on two-dimensional regularized plane projection. The encoding method includes: acquiring reference-frame point cloud data and current-frame point cloud data; performing two-dimensional regularized plane projection on the reference-frame point cloud data to obtain a first two-dimensional projection plane structure, and performing two-dimensional regularized plane projection on the current-frame point cloud data to obtain a second two-dimensional projection plane structure; performing motion compensation and encoding on one or more pieces of first two-dimensional graphic information obtained based on the first two-dimensional projection plane structure and one or more pieces of second two-dimensional graphic information obtained based on the second two-dimensional projection plane structure to obtain bit stream information of the current-frame point cloud data; and repeating the foregoing steps until all frames of a point cloud sequence are encoded. The present invention can fully utilize spatio-temporal correlation of a point cloud sequence for encoding the point cloud sequence, which reduces temporal and spatial redundancy of the point cloud sequence and improves encoding efficiency.