Quantized Depth Point Cloud Compression via Patch Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in efficiently compressing and decompressing large point clouds, which are generated by sensors like LIDAR systems, due to the high volume of spatial and attribute information they contain.
Innovation Solution
A system that uses an encoder to compress point clouds by dividing them into patches, generating patch images with spatial and depth information, and encoding these images using video compression techniques. The decoder then reconstructs the point cloud from the compressed images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If point clouds are captured with high spatial and attribute information density, then measurement precision and information completeness are improved, but data volume and storage costs increase significantly
Solution Approach 1:
The point cloud is divided into multiple patches based on spatial proximity and attribute similarity. Each patch is processed independently through projection and quantization, allowing localized optimization of compression parameters while preserving overall data fidelity. This segmentation enables efficient processing of large point clouds by breaking them into manageable units.
Solution Approach 2:
The patent applies quantization to transform continuous depth values into discrete depth levels, and projects 3D point coordinates onto 2D image planes. These parameter transformations reduce the precision requirements for storing spatial information while maintaining perceptual quality, thereby significantly reducing data volume without complete loss of measurement precision.
2Productivity
If point clouds are compressed using traditional methods, then data transmission efficiency is improved, but reconstruction quality and spatial information integrity deteriorate
Solution Approach 1:
The patent projects 3D point cloud data onto 2D image planes, transforming the data structure from three-dimensional coordinates to two-dimensional pixel representations. This dimensionality reduction enables the application of efficient 2D image compression algorithms while preserving essential spatial relationships through the projection geometry and depth encoding.
Solution Approach 2:
The patent introduces an intermediate representation where point clouds are converted to projected images with associated depth maps. This intermediate form serves as a bridge between the original 3D data and the compressed output, allowing sophisticated compression techniques to be applied while maintaining the ability to reconstruct accurate spatial information through inverse projection.
3Manufacturing precision
If quantization depth levels are increased, then reconstruction accuracy is improved, but compression ratio and storage efficiency worsen
Solution Approach 1:
The patent applies different quantization depth levels to different regions of the point cloud based on their importance and complexity. Critical regions with high geometric detail or semantic importance receive finer quantization (higher depth levels), while less important regions use coarser quantization. This local adaptation optimizes the balance between reconstruction accuracy and storage efficiency by allocating bits where they provide the most value.
Data Source
AI summary
A system comprises an encoder configured to compress attribute information and/or spatial information for a point cloud and/or a decoder configured to decompress compressed attribute and/or spatial information for the point cloud. The encoder is configured to convert a point cloud into an image based representation. The encoder packs patch images into an image frame and fills empty spaces in the image frame with a padding. The encoder is also configured to determine quantized minimum depths and/or maximum depths patch images in the image frames, wherein depth information is signaled relative to the quantized minimum depth.