3D Point Cloud Encoding via Graph Dictionary Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current 3D point cloud compression methods are inefficient due to high computational complexity, limited compression performance, and failure to adaptively represent complex spatially irregular signals, especially in non-Euclidean spaces, leading to excessive resource requirements and distortion in transmission and storage.
Innovation Solution
A graph dictionary learning method for 3D point cloud compression, which involves voxelization, graph dictionary construction using a nonlinear p-Laplacian operator, and iterative optimization to obtain a graph dictionary for encoding and decoding, allowing for adaptive sparse representation and efficient compression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If graph Fourier transform is used to decorrelate point cloud attributes, then compression performance is improved, but computational complexity becomes very high due to eigenvalue decomposition
Solution Approach 1:
The patent replaces the expensive graph Fourier transform (requiring eigenvalue decomposition) with a cheaper discrete cosine transform (DCT) that can be computed efficiently without decomposition. This substitutes a computationally intensive transformation with a more efficient one that achieves similar decorrelation goals.
Solution Approach 2:
The patent changes the transformation basis from graph Fourier basis (eigenvectors of Laplacian matrix) to DCT basis functions. This parameter change in the transformation method dramatically reduces computational complexity while maintaining the ability to decorrelate spatially correlated point cloud attributes.
2Reliability
If pre-defined transform based on strict mathematical expressions is used, then geometry information is exploited, but adaptability to complex spatially irregular multidimensional signals is lost
Solution Approach 1:
The patent introduces a dynamic adaptive mechanism where the transform basis is not fixed but can be adapted to the specific characteristics of the point cloud data. The method dynamically selects and adjusts transform parameters based on the spatial distribution and correlation properties of the actual data, enabling adaptability to complex irregular signals while maintaining geometric awareness.
Solution Approach 2:
The patent segments the point cloud data into different spatial regions or blocks and applies adaptive transforms to each segment independently. This allows different transform parameters to be used for different regions, improving adaptability to local signal characteristics while preserving overall geometric structure.
3Shape
If point cloud data is compressed without considering statistical characteristics, then geometry structure is preserved, but compression efficiency of attribute signals is reduced
Solution Approach 1:
The patent incorporates statistical analysis of point cloud attributes as feedback to guide the compression process. By analyzing the statistical characteristics (mean, variance, correlation) of the attribute signals, the system adjusts transform parameters and quantization strategies to optimize compression efficiency while maintaining geometric fidelity.
Solution Approach 2:
The patent performs preliminary statistical analysis and preprocessing of the point cloud attribute signals before applying the main compression transform. This preliminary action includes computing statistical moments, identifying correlation patterns, and pre-processing the data to enhance the effectiveness of subsequent compression operations.
Data Source
AI summary
A graph dictionary learning method for a 3D point cloud comprises: obtaining N point clouds to form training dataset; performing voxelization process on the point cloud data to obtain voxelized point cloud data of the training dataset; performing voxel block division on the point cloud data of the training dataset, selecting a plurality of voxel blocks as the training dataset, and constructing a graph dictionary learning model according to the training dataset; and performing iterative optimization on the graph dictionary learning objective function to obtain a graph dictionary for encoding and decoding a 3D point cloud signal. The present disclosure effectively uses the spatial correlation between point cloud signals to near-optimally remove the redundancy among point cloud signals.