3D Point Cloud Compression With Shape-Adaptive Transforms

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for compressing 3D point clouds, which often contain 'holes' or unoccupied positions, are inefficient as they are based on block-based techniques designed for images and videos where all positions are occupied, leading to difficulties in applying prediction and transforms effectively.

Innovation Solution

A method for compressing 3D point clouds by partitioning them into blocks, using projections and modified shape-adaptive transforms that accommodate 'holes' or unoccupied positions, and employing preprocessing to arrange points on a uniform grid, followed by intra prediction and either 3D shape-adaptive DCT or graph transforms.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If block-based compression methods designed for images and videos are applied to 3D point clouds, then compression efficiency can be improved, but the methods fail to handle unoccupied positions or 'holes' in the point cloud blocks

Engineering Contradiction:
Improvecompression efficiencyVSAvoidability to handle unoccupied positions
Core Design Contradiction:
ProductivityVSAdaptability or versatility

Solution Approach 1:

The patent applies different processing strategies to different regions within a block based on occupancy status. Occupied positions undergo prediction and transform operations, while unoccupied positions are handled separately through occupancy mapping. This local differentiation allows the system to maintain compression efficiency for occupied regions while properly representing the sparse structure of point clouds.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent segments the block processing into two distinct components: occupancy information (which positions are filled) and attribute information (values at occupied positions). This segmentation allows the use of traditional block-based compression for attributes while separately encoding occupancy patterns, thereby resolving the contradiction between compression efficiency and handling unoccupied positions.

Inventive Principle:
Principle #1Segmentation

2Loss of energy

If traditional prediction and transform operations are applied to all positions in a block, then compression can be achieved, but computational resources are wasted on unoccupied positions

Engineering Contradiction:
Improvecomputational resource efficiencyVSAvoidcompression performance
Core Design Contradiction:
Loss of energyVSProductivity

Solution Approach 1:

The patent applies prediction and transform operations only to occupied positions rather than all positions in the block. By performing partial action only where needed (at occupied positions), the system avoids wasting computational resources on unoccupied positions while maintaining compression performance through selective processing.

Inventive Principle:
Principle #16Partial or excessive action

3Reliability

If the point cloud is represented with high detail to maintain quality, then the data size increases significantly, but compression becomes more difficult

Engineering Contradiction:
Improvepoint cloud qualityVSAvoiddata size
Core Design Contradiction:
ReliabilityVSQuantity of substance

Solution Approach 1:

The patent transforms the point cloud data from spatial coordinates to frequency domain representations through transform operations (such as DCT or wavelet transforms). This parameter change allows the data to be compressed by concentrating energy in fewer coefficients, thereby reducing data size while maintaining reconstruction quality through selective coefficient encoding.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3405928B1Method for compressing point cloud
Publication Date: 2025.07.16 MITSUBISHI ELECTRIC CORP
  • EP3405928B1 patent drawingFigure 1
  • EP3405928B1 patent drawingFigure 2A~2B
  • EP3405928B1 patent drawingFigure 3A~3B

AI summary

A method compresses a point cloud composed of a plurality of points in a three-dimensional (3D) space by first acquiring the point cloud with a sensor, wherein each point is associated with a 3D coordinate and at least one attribute. The point cloud is partitioned into an array of 3D blocks of elements, wherein some of the elements in the 3D blocks have missing points. For each 3D block, attribute values for the 3D block are predicted based on the attribute values of neighboring 3D blocks, resulting in a 3D residual block. A 3D transform is applied to each 3D residual block using locations of occupied elements to produce transform coefficients, wherein the transform coefficients have a magnitude and sign. The transform coefficients are entropy encoded according the magnitudes and sign bits to produce a bitstream.