Point Cloud Coding With Bottom-Up Occupancy Probability Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing learning-based octree coding methods for point cloud compression suffer from inaccurate occupancy probability estimation and high computational complexity due to reliance on parent octree level information, leading to inefficient bit usage and processing overhead.
Innovation Solution
A bottom-up approach for encoding and decoding point cloud data using finer level details, incorporating feature extraction and aggregation from child voxels, and employing hyperprior encoding for enhanced probability estimation and arithmetic coding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If learning-based octree coding methods rely on parent octree level information for occupancy probability estimation, then the coding process can be simplified, but the accuracy of occupancy probability estimation deteriorates and computational complexity increases
Solution Approach 1:
The patent inverts the traditional top-down approach by adopting a bottom-up approach where occupancy probability is estimated from child voxels rather than parent voxels. The feature aggregation module aggregates features from child voxels to estimate occupancy probability for the current voxel, fundamentally reversing the information flow direction and resolving the contradiction between simplicity and accuracy.
Solution Approach 2:
The patent introduces a new dimension of information flow by aggregating features from child voxels (finer level details) to estimate occupancy at the current voxel level. This dimensional shift from parent-to-child information flow enables more accurate occupancy probability estimation while maintaining coding efficiency through the hierarchical structure.
2Device complexity
If learning-based octree coding methods use parent octree level information for feature extraction, then the feature extraction process is simplified, but computational complexity increases due to processing overhead
Solution Approach 1:
The patent inverts the traditional feature extraction direction by extracting features from child voxels rather than parent voxels. The feature aggregation module processes child voxel features to generate the current voxel's feature representation, which reduces computational complexity by avoiding the need to process entire parent octree levels while maintaining feature extraction accuracy.
Solution Approach 2:
The patent applies partial action by only processing and aggregating features from child voxels that are necessary for the current voxel's occupancy estimation, rather than processing the entire parent octree level. This selective processing reduces computational overhead while maintaining sufficient feature representation accuracy.
3Ease of manufacture
If existing methods use traditional arithmetic coding for occupancy encoding, then the encoding process is straightforward, but bit usage efficiency deteriorates
Solution Approach 1:
The patent changes the parameters of the arithmetic coding process by using occupancy probability estimates derived from child voxel features rather than traditional parent voxel information. This parameter change enables more efficient bit allocation during encoding, improving bit usage efficiency while maintaining encoding simplicity through the automated probability estimation process.
Data Source
AI summary
In one implementation, point cloud data for a point cloud is decoded. The decoder obtains features representing voxels in a tree structure, where feature for a current voxel is representative of at least a set of voxels that are still to be reconstructed. The decoder then determines an occupancy probability of the current voxel based on the feature, and decodes occupancy information of voxels in the tree structure, where whether a current voxel is occupied or not is decoded based on the occupancy probability for the current voxel. The point cloud can be reconstructed based on the occupancy information. On the encoder side, the feature for the current voxel is obtained from the voxels that are still to be encoded and encoded into a bitstream.