Point Cloud Attribute Prediction with Shifted Reference Groups
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current point cloud coding technologies limit the performance of attribute prediction by restricting the obtaining of neighboring points based on a current point, reducing the efficiency of point cloud attribute prediction.
Innovation Solution
The method involves obtaining an alternative point set from multiple point cloud groups with identical coordinate code words after shifting by a grouping shift bit number, predicting attribute values using prediction reference points, and decoding a code stream to determine attribute reconstruction values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If neighboring points are obtained based on a current point with a limited spatial structure, then the complexity of obtaining neighboring points is reduced, but the performance of point cloud attribute prediction deteriorates
Solution Approach 1:
The patent extends the prediction scope from local spatial neighbors to global alternative point sets by introducing a new dimension of selection based on attribute similarity. Instead of being constrained to spatial proximity, the system now considers points from any location in the point cloud that share similar attributes, effectively adding an attribute-based dimension to the neighbor selection process.
Solution Approach 2:
The patent changes the selection criteria from spatial coordinates to attribute values. By using attribute similarity as the primary selection parameter instead of spatial distance, the system identifies alternative points that may be located anywhere in the point cloud, not just in the immediate spatial neighborhood, thereby improving prediction accuracy without increasing computational complexity.
2Measurement precision
If alternative point sets are obtained from multiple point cloud groups considering spatial correlations, then the accuracy of attribute prediction is improved, but the complexity of the prediction system increases
Solution Approach 1:
The patent performs preliminary grouping of point cloud points into multiple point cloud groups before the prediction process. By pre-organizing points into groups with similar coordinate code words, the system prepares alternative point sets in advance, making the actual prediction process simpler and reducing the computational burden during runtime.
Solution Approach 2:
The patent creates alternative point sets that are copies or representations of existing point cloud groups. These alternative sets serve as reference models for prediction, allowing the system to leverage proven spatial correlations from existing groups without having to re-analyze all point cloud data from scratch for each prediction task.
Data Source
AI summary
In a method for point cloud prediction processing, an alternative point set of a target point cloud group is obtained from M point cloud groups. The M point cloud groups include the target point cloud group. Coordinate code words of point cloud points in each of the M point cloud groups are identical after shifting by a grouping shift bit number corresponding to the respective point cloud group. Prediction reference points associated with a target point cloud point of the point cloud points in the target point cloud group are obtained from the alternative point set. The target point cloud point is predicted by processing circuitry based on the prediction reference points, to obtain a target predicted attribute value of the target point cloud point.