Point Cloud Attribute Encoding Using Neighboring Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current point cloud attribute compression methods are inefficient, leading to challenges in storage and transmission due to the large amount of data generated by 3D scanning technologies, and there is a need for improved schemes to handle this data effectively.
Innovation Solution
The proposed solution involves an encoding and decoding method that calculates a first parameter based on neighboring points in a point cloud, using a preset prediction mode to determine a prediction value for attribute encoding and decoding, and quantizing residual values for efficient data representation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional point cloud attribute compression methods are used, then the compression process can be performed, but the compression efficiency is insufficient and data redundancy remains high
Solution Approach 1:
The patent applies preliminary action by performing prediction operations before encoding. Specifically, it uses neighboring point attributes to predict current point attributes, and performs lifting transforms on prediction residuals before final encoding. This preliminary prediction and transform process reduces the amount of data that needs to be encoded, thereby improving compression efficiency while reducing data redundancy.
Solution Approach 2:
The patent employs parameter changes by introducing multiple prediction modes (first prediction mode using neighboring point attributes, second prediction mode using lifting transforms) and dynamically selecting among them based on rate-distortion optimization. It also changes the parameter representation by using quantization parameters and transform coefficients instead of raw attribute values, which improves compression efficiency and reduces redundant data storage.
2Measurement precision
If more detailed attribute data is retained, then measurement precision is improved, but storage requirements and transmission complexity increase
Solution Approach 1:
The patent extracts only the essential and non-redundant attribute information by using prediction-based compression. It separates the predictable components (captured by neighboring point attributes and lifting transforms) from the residual information that needs to be stored. This extraction process maintains measurement precision for the residual data while significantly reducing the overall data volume that needs to be stored and transmitted.
Solution Approach 2:
The patent applies local quality by using different prediction modes and transform strategies for different regions or contexts within the point cloud data. It adapts the compression approach locally based on the characteristics of neighboring points and rate-distortion considerations, ensuring that precision is maintained where needed while reducing data volume in regions where prediction is effective.
Data Source
AI summary
An encoding method includes following operations. Neighbouring points of a current point in a point cloud to be encoded are determined, and a first parameter is calculated according to the neighbouring points. The first parameter is a difference between a maximum value and a minimum value among reconstructed values of first attributes of the neighbouring points. In response to the first parameter being less than a threshold, a prediction value of the first attribute of the current point is determined by using a preset first prediction mode. A difference between an original value of the first attribute of the current point and the prediction value is calculated as a residual value of the first attribute of the current point. The residual value subjected to quantization is encoded. Identification information of the first prediction mode is signalled, where the identification information is used for indicating a prediction mode.