Point Cloud Geometry Padding Using Compressed Occupancy Map
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current point cloud compression methods, such as TMC2, use 2D dilation for image padding, which can cause geometric distortions due to the incorporation of empty positions into the reconstructed point cloud, as they only consider 2D information.
Innovation Solution
The method involves using a compressed occupancy map to perform 3D geometry padding, where reconstructed points are selected based on 3D information to minimize distance from the original point cloud, thereby reducing geometric distortions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If 2D dilation is used for image padding, then the padding process is simple and fast, but geometric distortions occur due to incorporation of empty positions into the reconstructed point cloud
Solution Approach 1:
The patent transitions from 2D image space padding to 3D point cloud space padding. By performing padding operations in the 3D reconstructed point cloud domain rather than in the 2D geometry image domain, the method preserves geometric accuracy while maintaining efficiency. The 3D padding uses the spatial relationships and depth information inherently available in the point cloud representation to fill empty positions without causing geometric distortions.
Solution Approach 2:
The patent introduces the compressed occupancy map as an intermediary tool to guide the 3D padding process. The occupancy map identifies empty positions in the reconstructed point cloud, which then serves as a mask for the 3D padding operation. This intermediary structure enables the padding process to target only the necessary empty positions in 3D space, improving both accuracy and efficiency.
2Measurement precision
If uncompressed occupancy map is used for padding, then all empty positions are visible and can be padded, but the padding process becomes complex and computationally intensive
Solution Approach 1:
The patent changes the state of the occupancy map from uncompressed to compressed form. The compressed occupancy map retains the essential information about empty positions while reducing the data volume and processing complexity. This parameter change (compression ratio) allows the system to maintain accurate empty position identification while significantly reducing the computational burden of the padding process.
3Ease of manufacture
If 2D dilation padding is applied, then the padding algorithm is simple to implement, but the reconstructed point cloud contains geometric distortions
Solution Approach 1:
The patent moves the padding operation from 2D image space to 3D point cloud space. In 3D space, the padding algorithm can utilize depth information and spatial relationships to place padding points accurately, avoiding the geometric distortions that occur when 2D dilation is projected back to 3D. The simplicity is maintained by using the same dilation concept but applying it in the appropriate 3D domain where it produces accurate results.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
A method of point cloud geometry padding is described herein. The method searches for a reconstruction point in a compressed occupancy map to perform the geometry padding instead of using an uncompressed occupancy map.