Adaptive Point Cloud Compression via Neighborhood Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Point cloud files, generated by sensors like LIDAR systems, are large and costly to store and transmit, limiting their use in real-time applications due to high storage and network resource requirements.
Innovation Solution
A system that compresses attribute information of point clouds using an encoder that assigns attribute values and determines attribute correction values based on neighboring points, allowing for efficient storage and transmission, and decompresses the data using a decoder that adjusts predicted attribute values with these corrections.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If point cloud data is stored and transmitted in raw format, then data quality and completeness are maintained, but storage space and network resources are excessively consumed
Solution Approach 1:
The patent extracts only the necessary attribute information from the complete point cloud data. By identifying and transmitting only the essential attributes (such as color, intensity, reflectivity) rather than all possible attributes, the system reduces storage and transmission requirements while maintaining the quality needed for real-time applications.
Solution Approach 2:
The system changes the parameters of attribute information by selecting specific attributes to compress and transmit. Instead of transmitting all attribute data, the system modifies the data representation to include only the most relevant parameters, thereby reducing data volume while preserving essential information for applications like augmented reality and virtual reality.
2Productivity
If all attribute information is transmitted for every point, then data completeness is maintained, but transmission time and network bandwidth are excessively consumed
Solution Approach 1:
The patent segments the attribute information transmission by dividing points into different categories (e.g., foreground objects, background elements, occluded regions). Different segmentation strategies are applied to different point groups, allowing critical points to be transmitted with full attribute data while non-critical points use reduced attribute sets, thereby enabling real-time processing without excessive transmission time.
Solution Approach 2:
The system applies partial action by transmitting attribute information selectively rather than completely for all points. Instead of transmitting all possible attributes for every point, the system transmits only the necessary attributes for points that are most relevant to the current application task, reducing transmission time while maintaining sufficient data for real-time rendering and processing.
3Measurement precision
If comprehensive attribute data is stored, then data accuracy is maintained, but storage cost and computational overhead increase
Solution Approach 1:
The patent applies local quality by differentiating the treatment of attribute information based on spatial location and point importance. Critical points (such as those representing foreground objects or salient features) maintain high measurement precision with comprehensive attributes, while non-critical points (background elements, occluded regions) use simplified attribute representations, thereby reducing overall system complexity while preserving accuracy where needed.
Solution Approach 2:
The system changes the parameter representation dynamically based on point characteristics. Instead of using a fixed set of attributes for all points, the system adjusts the attribute parameter set according to the specific needs of each point group, reducing computational overhead for points that don't require full attribute detail while maintaining measurement precision for points where it is critical.
Data Source
AI summary
A system comprises an encoder configured to compress attribute information for a point cloud and/or a decoder configured to decompress compressed attribute for the point cloud. To compress the attribute information, attribute values are predicted using one of a plurality of prediction strategies, wherein a selected prediction strategy is selected based at least in part on attribute variability of points in a neighborhood of points. A decoder follows a similar prediction process. Also, attribute correction values may be determined to correct predicted attribute values and may be used by a decoder to decompress a point cloud, wherein the decoder applies the same prediction strategy applied at the encoder.