Point Cloud Attribute Encoding Using Spatial Predictor Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current point cloud compression technologies, such as MPEG-I part 5 (V-PCC) and MPEG-I part 9 (G-PCC), are inefficient for sparse point clouds and require complex encoding processes, leading to suboptimal compression performance and high latency, especially when dealing with non-smooth patches from Lidar-acquired data.
Innovation Solution
A method for encoding and decoding point cloud attributes that selects a predictor list of points based on their relative geographic position, allowing for the inclusion of points farther away if they provide better attribute prediction, thereby improving prediction accuracy and reducing the bitstream size, using a processor and memory storage device to execute the encoding and decoding processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional compression methods (MPEG-I part 5 V-PCC or MPEG-I part 9 G-PCC) are used, then the encoding process is complex, but the compression performance is suboptimal and latency is high especially for non-smooth patches from Lidar data
Solution Approach 1:
The patent applies local quality by selecting predictor points based on their spatial relationship to the current point, creating locally adapted prediction models. The method determines a set of predictor points from the point cloud that are spatially related to the current point, allowing each point to be encoded using locally relevant predictors rather than a uniform global approach, thereby improving compression performance for non-smooth patches while maintaining manageable complexity through localized processing
Solution Approach 2:
The patent implements dynamics by making the prediction model adaptive to each point's local characteristics. The set of predictor points is dynamically selected based on the spatial distribution and relationships in the local neighborhood of each point being encoded, allowing the encoding process to adapt to varying point cloud densities and geometric structures without requiring a fixed complex encoding framework
2Measurement precision
If points farther away are included in the predictor list, then prediction accuracy improves, but the bitstream size increases
Solution Approach 1:
The patent applies parameter changes by optimizing the selection criteria for predictor points based on their spatial distribution and predictive capability. Instead of using a fixed number of nearest neighbors, the method dynamically determines the optimal set of predictor points by considering their spatial relationships and contribution to prediction accuracy, thereby achieving high prediction accuracy while minimizing the number of points that need to be encoded in the bitstream
Data Source
AI summary
A method for encoding, in a bitstream, point attributes associated to a point of a point cloud, includes: determining, for a point P to be encoded to the bitstream, a predictor list of k predictor points of the point cloud including k points of the point cloud nearest to the point P to be encoded, wherein the k points are selected according to their relative position to each other; and encoding the point attributes associated to the point P to be encoded by predictive encoding based on attributes of the predictor points of the predictor list.