Video Encoder Merge Candidate Selection for Screen Content
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High-Efficiency Video Coding (HEVC) faces challenges in reducing complexity and improving performance, particularly in motion vector difference (MVD) signaling and prediction modes for video coding, especially for screen content with repetitive patterns.
Innovation Solution
The method involves generating a set of Merge candidates with MMVD candidates by modifying motion vector expressions and conditionally signaling new merge tools, such as disabling certain prediction modes like CPR, MMVD, and triangular prediction units when no other Inter reference pictures are available, to simplify motion vector signaling and reduce encoder complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple prediction modes (CPR, MMVD, triangular prediction units) are enabled to improve coding efficiency, then compression performance is improved, but encoder complexity increases
Solution Approach 1:
The patent implements dynamic enabling/disabling of prediction modes (CPR, MMVD, triangular prediction units) based on reference picture availability. When no other Inter reference pictures are available, these modes are disabled; otherwise, they are enabled to improve compression performance. This dynamic adaptation resolves the contradiction by adjusting complexity according to actual coding conditions.
Solution Approach 2:
The patent changes the operational state (enabled/disabled) of prediction modes as a parameter based on reference picture availability. This parameter change allows the encoder to switch between high-complexity/high-performance mode and low-complexity/standard mode, resolving the contradiction between compression performance and encoder complexity.
2Productivity
If MMVD candidates are generated by modifying motion vector expressions to extend motion vector diversity, then coding efficiency is improved, but signaling complexity increases
Solution Approach 1:
The patent generates MMVD candidates by modifying motion vector expressions only when reference pictures are available, rather than always generating them. This partial action approach improves coding efficiency when beneficial while avoiding unnecessary signaling complexity when reference pictures are unavailable, resolving the contradiction between coding efficiency and signaling complexity.
3Device complexity
If prediction modes are conditionally disabled when no other Inter reference pictures are available, then encoder complexity is reduced, but coding efficiency may be limited
Solution Approach 1:
The patent dynamically adjusts the set of available prediction modes based on reference picture availability. When reference pictures are unavailable, modes are disabled to reduce complexity; when available, modes are enabled to maximize coding efficiency. This dynamic approach ensures optimal balance between complexity and efficiency for each coding scenario.
Solution Approach 2:
The patent changes the operational status of prediction modes as a controllable parameter based on reference picture availability. This parameter change allows the encoder to adapt its complexity level and corresponding coding efficiency based on the actual reference picture context, resolving the contradiction between reduced complexity and maintained efficiency.
Data Source
AI summary
A prediction method and apparatus for video coding include a set of Merge candidates having an MMVD (Merge mode with motion vector differences (MVD)) candidate are generated. When the current block has a block size belonging to at least one default block size, whether a target candidate in the set of Merge candidates is bi-prediction is determined after the set of Merge candidates is generated. If the target candidate is bi-prediction: the target candidate is replaced by a uni-prediction candidate; the uni-prediction candidate is included in the set of Merge candidates to generate a modified set of Merge candidates; and current motion information associated with the current block is encoded using the modified set of Merge candidates at the video encoder side or the current motion information associated with the current block is decoded at the video decoder side using the modified set of Merge candidates.