3D Point Cloud Inter-Predictive Encoding via Residual Translation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for compressing dynamic 3D point clouds in immersive and augmented reality systems require high bitrates for encoding rigid transformation matrices, particularly for translation vectors, which hinders efficient communication in bandwidth-limited networks.
Innovation Solution
The method encodes only the estimation error of the translation vector instead of the vector itself, using a geometric transformation that includes rotation and translation information, and represents the rotation transformation using Euler angles or unit quaternions for improved compression efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If rigid transformation matrices including full translation vectors are encoded, then accurate geometric transformation is achieved, but bitrate requirement increases significantly
Solution Approach 1:
The translation vector encoding is segmented into a predicted translation vector (derived from rotation matrix and reference block positions) and a residual correction vector. Only the residual correction vector is fully encoded, while the predicted portion is inferred by the decoder, thereby reducing the bitrate required for translation information while maintaining geometric transformation accuracy.
Solution Approach 2:
The decoder performs preliminary computation of the translation vector using the decoded rotation matrix and known reference block positions before applying the encoded residual correction. This preliminary action allows the system to encode only the difference (residual) rather than the complete translation vector, reducing bitrate while preserving accuracy.
2Productivity
If inter-frame redundancy is exploited via inter prediction, then compression performance improves, but complexity of encoding rigid transformation matrices increases
Solution Approach 1:
The encoding approach changes the parameters being encoded from complete rigid transformation matrices to only the residual translation correction vectors. This parameter change exploits inter-frame redundancy by assuming the majority of translation can be predicted from rotation and block positions, thereby improving compression performance while managing encoding complexity through selective encoding of only necessary correction data.
Data Source
AI summary
This method for inter-predictive encoding of a time-varying 3D point cloud including a series of successive frames divided in 3D blocks into at least one bitstream comprises encoding (20) 3D motion information including a geometric transformation comprising rotation information representative of a rotation transformation and translation information representative of a translation transformation, wherein the translation information comprises a vector ΔT representing an estimation error of the translation transformation.