Point Cloud Data Processing Scalable Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current point cloud compression techniques either lose original data by modeling it as a surface or rely on assumptions about data correlation, leading to suboptimal compression and artifacts, especially for arbitrarily distributed three-dimensional data.
Innovation Solution
The method involves defining cell boundaries in the point cloud, assigning unique indexes, and mapping coordinates to a local system, allowing for index-based processing that preserves the original distribution, using techniques like multidimensional reference gridding, sub-regional distance transform, and sub-band decomposition for scalable compression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of substance
If point cloud is modeled as a surface using triangle networks, then data compression is achieved, but the original point cloud data is lost and artifacts are introduced
Solution Approach 1:
The point cloud data is segmented into multiple levels of detail, where the original point cloud is divided into coarse and fine components. The coarse level provides a simplified representation for compression, while the fine level preserves detailed features, allowing reconstruction of the original data without loss.
Solution Approach 2:
The patent implements a nested structure where multiple levels of triangle networks are embedded within each other. Each level contains points from the previous level plus additional detail points, creating a hierarchical representation that preserves original data while enabling progressive compression at different quality levels.
2Loss of substance
If wavelet analysis is applied to time series of vectors, then compression is achieved for correlated data, but approximation errors are introduced and the technique is platform-dependent
Solution Approach 1:
The patent transforms the data representation by changing parameters from absolute coordinates to relative offsets and differences. This parameter transformation reduces the dynamic range and creates redundancy that can be exploited for compression while maintaining exact reconstruction, avoiding approximation errors inherent in wavelet analysis.
Data Source
AI summary
A point cloud contains data points having coordinates that locate each data point in a coordinate system. Cell boundaries are defined in the point cloud and a unique index is assigned to each cell. The indexes of the cells are assigned to any of data point contained in a cell having the same index. The index data points can be accessed in a manner analogous to rasterized or gridded data while maintaining the arbitrary or random distribution associated with point clouds. The indexed point cloud can be further processed to achieve a scalable compression level, either losslessly by increasing redundancy in the point cloud data and/or by separating the point cloud data into relevancy bands, or by conventional lossy techniques, such as by exploiting the redundancies created in the point cloud data to represent the point cloud in a finite number of symbols.