Point Cloud Attribute Prediction Through Adaptive Neighbor Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting point cloud attributes based on preset neighbor point parameters are undiversified and inefficient, leading to low decoding efficiency.
Innovation Solution
Determine neighbor point parameters based on a target association relationship, including the quantity of candidate points, neighbor points, and distance between points, to enrich prediction methods and improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If preset neighbor point parameters are used for prediction, then the prediction process is simple, but the prediction efficiency and diversity are low
Solution Approach 1:
The patent applies dynamics by making the neighbor point parameters adaptive rather than fixed. The parameters (quantity of candidate points, quantity of neighbor points, preset distance) are dynamically determined based on the spatial parameters of the point cloud, allowing the prediction method to adjust to different data characteristics and improve decoding efficiency without overwhelming complexity
Solution Approach 2:
The patent changes parameters by establishing a target association relationship that links neighbor point parameters to spatial parameters. This allows the prediction method to utilize spatial parameter information (such as point cloud density, distribution characteristics) to determine appropriate neighbor parameters, thereby enriching the prediction diversity and improving efficiency through parameter adaptation
Data Source
AI summary
A method includes: decoding a point cloud code stream to obtain decoded points in a point cloud; determining a target association relationship, the target association relationship including at least one association relationship of a preset association relationship between neighbor point parameters and a preset association relationship between a space parameter of the point cloud and the neighbor point parameters; obtaining the neighbor point parameters according to the target association relationship; selecting at least one neighbor point of a current point from the decoded points according to at least one of a quantity N of candidate points, a quantity K of neighbor points, and a preset distance d between the neighbor point and the current point in the neighbor point parameters; and determining a predicted value of attribute information of the current point according to attribute information of the at least one neighbor point of the current point.