Volumetric Video Point Cloud Compression via Depth Image Anchoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Volumetric video content captured by 3D cameras generates vast amounts of data, requiring significant bandwidth for storage and transmission, and existing compression methods are inefficient for highly complex and sparse point clouds.
Innovation Solution
The method involves encoding point clouds using Successive Refinement of Bounding Volumes with Outer Surface Anchor Points, which includes generating bounding volumes, encoding depth images, determining occupancy status, detecting anchoring points, and encoding missing elements using anchored chain codes, to efficiently compress and reconstruct 3D geometry.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If volumetric video content is captured using multiple 3D cameras, then the field of view and immersion experience are improved, but the data volume and bandwidth requirements increase significantly
Solution Approach 1:
The point cloud data is segmented by generating bounding volumes that divide the 3D space into manageable regions. Each bounding volume encloses a specific portion of the point cloud, allowing the data to be processed and transmitted in smaller, organized segments rather than as a single large dataset.
Solution Approach 2:
The patent transforms the 3D point cloud data into 2D depth images by projecting points onto image planes. This dimensionality reduction from 3D to 2D significantly compresses the data volume while preserving the essential geometric information needed for reconstruction.
2Ease of manufacture
If traditional compression methods are used for point cloud data, then implementation is straightforward, but compression efficiency is insufficient for highly complex and sparse point clouds
Solution Approach 1:
Before compressing the point cloud data, the patent performs preliminary organization by generating bounding volumes and depth images. This preprocessing step structures the data in a way that enables more efficient compression algorithms to be applied subsequently, improving overall compression efficiency.
Solution Approach 2:
The patent introduces depth images as an intermediary representation between the original point cloud and the compressed data. These depth images serve as a mediator that captures the essential geometric information in a compact 2D format, facilitating more efficient compression than direct point cloud encoding.
3Manufacturing precision
If all points in the point cloud are encoded in full detail, then reconstruction accuracy is maximized, but the encoding time and computational resources increase
Solution Approach 1:
The patent applies different encoding strategies to different regions of the point cloud based on their importance. Bounding volumes are generated to enclose significant portions of the point cloud, and depth images are created for these regions, allowing focused processing on important areas while reducing detail in less critical regions.
Solution Approach 2:
Instead of encoding all points with equal detail, the patent uses partial action by selectively encoding only the bounding volumes and depth images that capture the essential structure. This partial encoding approach achieves sufficient reconstruction accuracy for many applications while dramatically reducing encoding time and computational resources.
Data Source
AI summary
Encoding a two depth images of points of a bounded volume (730); determining a first section across a second axis between the depth images to generate a feasible occupancy image and a list of pixels with unknown occupancy status (740); encoding a binary sequence according to a true occupancy image, wherein the binary sequence transmits the occupancy status for the list of pixels with unknown occupancy status (750); detecting anchoring points from the binary sequence, and encoding missing elements between the anchoring points (760); determining a first section across a third axis from the bounding volume reconstructed so far and create a list of pixels with unknown status (780); encoding a binary sequence according to the true occupancy image the sequence conveying the occupancy status of the list of pixels with unknown status (785); and detecting the anchor points from the sequence and corresponding list of pixels and encoding missing elements between the anchoring points (795).