3D Point Cloud Encoding via Adaptive Sign Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing point cloud compression techniques, such as PCC, struggle to improve encoding efficiency for point groups distributed along the surface of objects, as they do not consider the shape of the object's surface or boundaries, leading to inefficient encoding.
Innovation Solution
An encoding device that acquires three-dimensional data and divides the parent space into child spaces, dynamically allocating a sign indicating point presence based on adjacent child spaces, allowing for the selection of an appropriate arithmetic encoding table to optimize encoding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional PCC encoding is used that divides parent space into child spaces and allocates signs based on point presence, then the encoding process is simple and follows a regular octree structure, but encoding efficiency is poor for point groups distributed along object surfaces because the method does not consider surface shape or boundaries
Solution Approach 1:
The patent applies local quality by making the sign allocation process adaptive to local surface characteristics. Instead of uniformly applying the same encoding rules across all child spaces, the method dynamically determines sign allocation based on whether adjacent child spaces contain points, thereby adapting to local surface geometry and boundaries. This allows the encoding to be more efficient in regions where points are concentrated along surfaces while maintaining simplicity in other regions.
Solution Approach 2:
The patent introduces dynamics by making the sign allocation process flexible rather than static. The allocation of signs to child spaces is dynamically adjusted based on the presence of points in adjacent child spaces. This dynamic adaptation allows the encoding method to respond to varying surface geometries and point distributions, improving encoding efficiency without requiring a completely complex restructuring of the octree framework.
2Quantity of substance
If signs are allocated to all child spaces regardless of point presence, then the encoding table is simple and uniform, but the code amount increases and compression performance deteriorates
Solution Approach 1:
The patent applies the taking out principle by selectively allocating signs only to child spaces that are relevant to the surface representation. Instead of uniformly allocating signs to all eight child spaces of each parent space, the method extracts and allocates signs only to those child spaces that contain points or are adjacent to child spaces containing points. This reduces the number of signs that need to be encoded, thereby reducing the code amount and improving compression performance.
Solution Approach 2:
The patent changes the parameter of sign allocation from a fixed uniform pattern to a variable pattern based on point presence in adjacent child spaces. By making the sign allocation dependent on the presence of points in neighboring regions, the method adapts the allocation pattern to match the actual data distribution. This parameter change allows the encoding to be more compact by avoiding allocation of signs to child spaces that would otherwise be empty or redundant.
Data Source
AI summary
An encoding device includes a division unit that acquires three-dimensional data representing positions of a plurality of points distributed along a surface of an object in a three-dimensional space and divides a parent space including the points in the three-dimensional space into a plurality of child spaces and an encoding unit that changes, based on a position of a target space, which is the child space, to which a sign representing whether the points are included is allocated, according to whether the points are included in a first child space adjacent to the target space, processing for allocating the sign to the target space and a second child space adjacent to the target space.