Adaptive Depth Prediction Lists for Discontinuous Point Clouds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for coding and decoding point cloud depth information fail to consider scenario discontinuity, leading to low prediction accuracy, numerous outliers, and inefficient coding due to reliance on neighboring point predictions.
Innovation Solution
Create an adaptive prediction list for point cloud depth information using multiple candidate lists, including values from same and different lasers, historical points, and priori information, to select optimal predicted values and reduce residuals and outliers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If depth information of neighboring points is used for prediction, then coding process is simple, but prediction accuracy is low and outliers increase due to scenario discontinuity
Solution Approach 1:
The patent segments the prediction process by dividing candidate prediction values into multiple categories (same laser line points, adjacent laser line points, historical prediction values) and selectively applying different prediction sources based on scenario continuity detection. This allows the system to handle discontinuous scenarios by switching between different prediction segments, improving accuracy without overwhelming complexity.
Solution Approach 2:
The patent introduces dynamic adaptation by continuously updating the prediction list based on detected scenario continuity. When discontinuity is detected, the system dynamically adjusts which prediction sources are used and in what order, making the prediction process adaptive to changing scenarios rather than relying on fixed neighboring point assumptions.
2Measurement precision
If adaptive prediction list with multiple candidate lists is created, then prediction accuracy improves, but device complexity increases
Solution Approach 1:
The complex prediction list is segmented into multiple specialized candidate lists (same laser line, adjacent laser line, historical predictions), each serving a specific prediction purpose. This segmentation allows the system to manage complexity by organizing prediction sources into manageable, functionally-distinct groups rather than a single undifferentiated list.
Solution Approach 2:
Different parts of the prediction list are assigned different qualities and priorities based on their reliability for specific scenarios. The system assigns higher weight to certain candidate lists based on local scenario characteristics (e.g., preferring same-laser-line points for continuous surfaces, historical predictions for discontinuous regions), optimizing prediction quality without uniform complexity.
3Device complexity
If fixed prediction mode list is used, then device complexity is low, but adaptability to different scenarios is poor
Solution Approach 1:
The patent transforms the static fixed prediction mode list into a dynamic adaptive prediction list that automatically adjusts its composition and ordering based on detected scenario continuity. This dynamic adjustment allows the system to adapt to different scenarios (continuous surfaces, discontinuous regions, various geometries) without requiring complex manual mode selection or multiple fixed mode lists.
Solution Approach 2:
The prediction list automatically updates and reorganizes itself based on feedback from prediction accuracy and scenario continuity detection, without requiring external intervention or complex control mechanisms. The system self-adjusts by promoting successful prediction sources and demoting unsuccessful ones, providing adaptability through self-service rather than complex external management.
Data Source
AI summary
A method for coding includes: obtaining original point cloud data; creating an adaptive prediction list of the depth information of the point cloud; and predictively coding the depth information of the point cloud based on the adaptive prediction list, to obtain codestream information. According to the present invention, an adaptively updated prediction list of the depth information is created, and an optimal prediction mode is selected from the list to predict the depth information of the point cloud.