Inter Prediction Merge List Redundancy Check
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding technologies face challenges in efficiently processing next-generation video contents with high spatial resolution and high frame rates due to increased memory storage and processing requirements, particularly in managing redundancy in motion vector prediction.
Innovation Solution
Implementing constraints on redundancy checks for adding candidates to the merge list (or AMVP list) to improve complexity and efficiency in video encoding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If redundancy checks are performed for adding HMVP candidates to merge list, then prediction accuracy is improved, but processing complexity increases
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting the redundancy check threshold based on block type, reference picture type, and motion vector characteristics. Different threshold parameters are used for different scenarios (e.g., higher thresholds for P-frames, lower for B-frames), allowing the system to maintain prediction accuracy while adapting processing complexity to specific content requirements
Solution Approach 2:
The patent implements local quality by applying different redundancy check strategies to different regions and block types within the video stream. Spatially, different thresholds are applied to different prediction blocks; temporally, different checks are applied to different frame types (P-frames vs B-frames). This localized approach ensures high prediction accuracy where needed while reducing complexity in less critical areas
2Measurement precision
If more HMVP candidates are added to merge list, then prediction accuracy is improved, but memory access rate increases
Solution Approach 1:
The patent applies partial action by performing redundancy checks selectively rather than universally. Redundancy checks are applied only when specific conditions are met (e.g., when motion vectors differ significantly, when block size exceeds threshold, or when reference picture types require it). This partial application maintains prediction accuracy for critical cases while avoiding unnecessary memory accesses in other cases
Solution Approach 2:
The patent segments the candidate selection process into multiple stages with different redundancy check intensities. The first stage performs basic filtering with minimal memory access, the second stage performs conditional redundancy checks on remaining candidates, and the third stage performs final validation only on top candidates. This segmented approach reduces overall memory access rate while maintaining final prediction accuracy
3Productivity
If pruning is applied to HMVP candidates, then processing speed is improved, but prediction accuracy may deteriorate
Solution Approach 1:
The patent applies preliminary action by performing preparatory filtering and classification of HMVP candidates before the main pruning operation. Candidates are pre-sorted by quality metrics and pre-filtered based on basic criteria (validity, reference picture availability). This preliminary organization allows the subsequent pruning operation to work more efficiently with already-organized data, maintaining accuracy by preserving high-quality candidates while removing only clearly inferior ones
Solution Approach 2:
The patent implements feedback mechanisms where pruning decisions are validated against actual prediction performance. The system monitors prediction accuracy after pruning and dynamically adjusts pruning thresholds and criteria. If accuracy degradation is detected, the system automatically relaxes pruning constraints. This closed-loop feedback ensures that processing speed improvements from pruning do not come at the cost of prediction accuracy
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The disclosure discloses a method for processing a video signal and an apparatus therefor. Specifically, the method of processing a video signal based on an inter prediction, comprising: configuring a merge list based on a neighboring block of a current block; adding a history based merge candidate included in a history based merge candidate list to the merge list when a number of a merge candidate included in the merge list is smaller than a first predetermined number; obtaining a merge index indicating a merge candidate used for an inter prediction of the current block within the merge list; and generating a prediction block of the current block based on motion information of a merge candidate indicated by the merge index, wherein a step of adding the history based merge candidate to the merge list comprises checking whether a second predetermined number of the history based merge candidate within the history based merge candidate list has the same motion information as the merge candidate included in the merge list.