Adaptive Transform Order for Point Cloud RAHT Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In Geometry-based Point Cloud Compression (G-PCC), the fixed transform order used in Region Adaptive Hierarchical Transform (RAHT) results in large information redundancy due to varying spatial distributions of different point clouds, leading to poor encoding and decoding efficiency.
Innovation Solution
A method that determines a transform order based on the point quantities corresponding to three-dimensional coordinate components by using a two-dimensional Morton code, prioritizing transform directions with significant features, and signaling this order into an attribute bitstream for efficient RAHT processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a fixed transform order is used in RAHT, then the encoding process is simple and fast, but large information redundancy occurs due to varying spatial distributions of different point clouds
Solution Approach 1:
The patent applies dynamics by making the transform order adaptive rather than fixed. The encoder determines the transform order dynamically based on the spatial distribution characteristics of each point cloud dataset, allowing the system to adapt to different spatial patterns while maintaining efficient encoding performance.
Solution Approach 2:
The patent changes the parameter of transform order from a fixed value to a variable determined by spatial distribution analysis. By calculating point quantities in different coordinate directions and selecting the order based on these calculations, the system optimizes the transform parameters to match the specific characteristics of each point cloud.
2Ease of operation
If a fixed transform order is used in RAHT, then the decoding process is simple, but encoding and decoding efficiency deteriorates due to poor adaptability to spatial distributions
Solution Approach 1:
The patent applies preliminary action by determining the optimal transform order during the encoding phase based on spatial distribution analysis. This pre-determined order is then signaled to the decoder, allowing the decoder to perform simple inverse transforms without needing to perform complex spatial analysis, thus maintaining decoding simplicity while improving efficiency.
Solution Approach 2:
The system uses feedback by analyzing the spatial distribution characteristics during encoding and using this information to determine the transform order. The determined order is communicated to the decoder through signaling, creating a feedback loop that optimizes the transform process based on actual data characteristics.
3Loss of information
If transform order is adapted to spatial distribution, then information redundancy is reduced, but the complexity of determining transform order increases
Solution Approach 1:
The patent applies segmentation by dividing the spatial distribution analysis into discrete steps: calculating point quantities in each coordinate direction (x, y, z), comparing these quantities, and selecting the transform order based on the comparison results. This segmented approach simplifies the overall complexity while achieving adaptability.
Solution Approach 2:
The patent changes the transform order parameter based on calculated spatial distribution parameters (point quantities in different directions). By using these calculated parameters to determine the transform order, the system achieves adaptability without requiring complex algorithms, as the determination process is based on straightforward quantity comparisons.
Data Source
AI summary
Provided in the embodiments of the present application are a transform method, inverse transform method, encoder, decoder and storage medium. The method comprises: determining a two-dimensional Morton code of an encoding point in a preset space of an encoding point cloud, wherein the two-dimensional Morton code is a Morton code corresponding to two-dimensional coordinates of the encoding point obtained by removing one of three-dimensional coordinates of the encoding point; performing right shift of the two-dimensional Morton code by K binary digits to determine point sets respectively corresponding to three-dimensional coordinate components; determining, from the point sets, point quantities of maximum included encoding points respectively corresponding to the three-dimensional coordinate components; determining, based on the point quantities, a transform order corresponding to the preset space; performing RAHT based on the transform order; and signaling the transform order and an attribute encoding result obtained after RAHT into a bitstream.