Point Cloud Encoding via Temporal Projection Consistency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for compressing and decoding point clouds, particularly dynamic ones, face inefficiencies due to occlusions, redundancy, and temporal instability, leading to high resource requirements and poor compression performance.
Innovation Solution
A method and apparatus for encoding and decoding point clouds by obtaining temporally successive pictures with consistent image arrangements, encoding information representative of unique projections for each image, and using these projections to encode and decode the point cloud, improving temporal consistency and compression efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple projections are used to handle occlusions in point cloud compression, then the completeness of geometry representation is improved, but the resource requirements and device complexity increase
Solution Approach 1:
The point cloud is divided into multiple patches, each projected onto a specific face of the cube. This segmentation allows comprehensive geometry representation through multiple projections while managing complexity by processing smaller, localized regions independently rather than handling the entire point cloud globally.
Solution Approach 2:
The patent transitions from traditional single-view or limited multi-view projections to a six-face cube projection system. By utilizing the three-dimensional cube structure with projections on all six faces, the system achieves complete geometry representation including occluded regions, effectively adding spatial dimensional coverage without proportionally increasing processing complexity.
2Productivity
If traditional projection methods are used for point cloud compression, then the implementation is simple, but the compression performance and temporal consistency are poor
Solution Approach 1:
The patent introduces dynamic patch selection and projection face assignment that adapts to the specific geometry and occlusion patterns of each point cloud dataset. Rather than using fixed projection methods, the system dynamically determines the optimal projection configuration, improving compression efficiency while managing complexity through adaptive algorithms.
Solution Approach 2:
The system varies projection parameters including which cube face receives which patch, the resolution allocation per face, and the temporal prediction parameters. By changing these parameters adaptively based on the point cloud characteristics, the system achieves superior compression performance without requiring fundamentally complex encoding infrastructure.
3Loss of information
If non-overlapping patches are used before projection, then the redundancy is reduced, but the projected partition boundaries become unsmooth and hard to code
Solution Approach 1:
The patent creates smooth transitions at patch boundaries by ensuring continuous depth values and consistent normal vectors across adjacent patches projected on the same or adjacent cube faces. This equipotential approach maintains geometric continuity at boundaries, making them smooth and easy to code while still achieving redundancy reduction through the non-overlapping patch structure.
Data Source
AI summary
Method and device for encoding a point cloud that represents a three-dimensional (3D) object. One or more groups of temporally successive pictures are obtained. Each picture of the one or more groups comprises a first set of images, the images being spatially arranged in a same manner in each picture of the one or more groups. A second set of projections is associated with the one or more groups, a unique projection being associated with each image in such a way that a same projection is associated with only one single image and all projections are associated with the images. A first information representative of the projections is encoded. The point cloud may then be encoded according to the obtained pictures. A corresponding method and device for decoding a bitstream that comprises data representative of the encoded picture(s) are also described.