Generalized Bi-Prediction Weighting for Block-Level Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in efficiently leveraging spatial and temporal redundancy in video content due to limitations in prediction methods, particularly in bi-prediction and weighted prediction, leading to suboptimal compression efficiency.
Innovation Solution
Implementing generalized bi-prediction (GBI) and weighted prediction (WP) tools in video encoding and decoding, allowing for flexible application and deactivation of these tools at a block level, and combining predictors using customizable weights and offsets to enhance prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional bi-prediction and weighted prediction methods are used, then video compression is achieved, but compression efficiency is suboptimal due to limitations in prediction methods
Solution Approach 1:
The patent applies parameter changes by introducing generalized bi-prediction weights (gbi_weights) that can take multiple discrete values (e.g., 0, 1, 2, 3) to control the contribution of different reference blocks. This allows the prediction system to adapt to various content characteristics by selecting appropriate weight parameters, thereby improving prediction accuracy and compression efficiency simultaneously
Solution Approach 2:
The patent implements dynamics by enabling flexible activation and deactivation of GBI and WP tools at block level based on content characteristics. The system dynamically selects prediction modes (GBI-only, WP-only, or combined) and adjusts weight parameters according to local picture content, allowing optimal adaptation to different video regions and improving overall compression efficiency
2Measurement precision
If multiple prediction tools (GBI and WP) are applied simultaneously, then prediction accuracy is enhanced, but computational complexity increases
Solution Approach 1:
The patent applies partial action by selectively activating GBI and WP tools only where needed based on content characteristics. Instead of applying both tools uniformly across all blocks, the system activates them partially based on local requirements, reducing unnecessary computational complexity while maintaining prediction precision where it matters most
Solution Approach 2:
The patent implements local quality by applying different prediction strategies (GBI, WP, or combined) to different blocks based on their specific characteristics. Each block can have customized prediction parameters and tool activation, allowing high prediction precision in complex regions while using simpler methods in homogeneous regions, thus balancing accuracy and complexity
3Productivity
If generalized bi-prediction is applied to all blocks, then compression efficiency improves, but processing overhead and complexity increase
Solution Approach 1:
The patent applies segmentation by dividing the picture into blocks and applying different prediction modes (GBI activation, WP activation, or both) to different segments based on content characteristics. This segmented approach allows compression efficiency improvements in suitable regions while maintaining processing simplicity in regions where GBI is not activated
4Adaptability or versatility
If block-level activation control is implemented, then adaptability to different content characteristics is improved, but syntax complexity and processing overhead increase
Solution Approach 1:
The patent implements universality by designing a unified syntax structure (gbi_enabled_flag, gbi_weights) that serves multiple functions: controlling GBI activation, selecting weight parameters, and coordinating with WP tool activation. This multi-functional syntax design improves content adaptability while minimizing the increase in syntax complexity through shared parameters and conditional activation logic
Data Source
AI summary
A decoding method is presented. At least one high level syntax element is decoded that indicates whether generalized bi-prediction applies for predicting blocks of a slice. A block is then decoded from said slice using generalized bi-prediction in the case where said at least one high level syntax element indicates to apply generalized bi-prediction.