Geometric Partition Boundary Prediction With Lower Pipeline Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding systems face challenges in achieving high compression efficiency due to limitations in predicting boundaries during geometric partitioning modes, leading to inefficiencies in encoding and decoding processes.
Innovation Solution
The method involves determining boundary candidates using geometric partitioning mode (GPM) and associating them with metrics for optimal encoding, and signaling the determined boundary candidate, along with prediction modes, to improve compression efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If geometric partitioning mode is used to improve compression efficiency, then compression ratio is improved, but boundary prediction accuracy deteriorates
Solution Approach 1:
The patent applies preliminary action by determining multiple boundary candidates before final encoding decisions are made. The encoder generates several possible boundary positions (e.g., horizontal, vertical, diagonal orientations) and evaluates them in advance using metrics such as template matching or rate-distortion optimization, then selects the optimal boundary for the geometric partitioning mode. This pre-evaluation process improves boundary prediction accuracy while maintaining compression efficiency.
2Measurement precision
If multiple boundary candidates are evaluated to improve prediction accuracy, then boundary prediction accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent applies partial action by evaluating only a limited set of predefined boundary candidates (e.g., 3-5 specific orientations) rather than exhaustively searching all possible boundary positions. This selective evaluation of partial candidates achieves sufficient prediction accuracy while significantly reducing computational complexity compared to a full search approach.
3Measurement precision
If advanced motion vector refinement and template matching are applied to improve coding performance, then coding performance is improved, but encoding time increases
Solution Approach 1:
The patent applies local quality by applying advanced motion vector refinement and template matching techniques selectively only to regions or blocks that benefit most from them, rather than uniformly across the entire video sequence. For example, complex template matching may be applied only to high-motion or high-importance blocks, while simpler methods are used elsewhere, thereby improving overall coding performance while limiting the increase in encoding time.
Data Source
AI summary
Prediction of a boundary used in geometric prediction modes is accomplished with embodiments that remove a requirement of reconstruction of samples of the current block, which increases pipeline latency. In one embodiment, the best boundary is predicted and added as a candidate to a list. In another embodiment, the candidates are ranked according to their boundary fitting score and an index is signaled to select the correct candidate at a corresponding decoder. In another embodiment, the fitting of the boundary is computed using a metric along the boundary sub-block. In another embodiment, the fitting of the boundary is computed using a metric outside the boundary area.