Sub-block Size Derivation for Affine Inter Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video compression techniques face challenges in reducing computational complexity, particularly with bidirectional affine inter-prediction, which is computationally intensive due to the large number of sub-blocks and motion vectors required, leading to increased processing demands and bandwidth usage.
Innovation Solution
Implementing sub-block size thresholds that differentiate between unidirectional and bidirectional affine inter-prediction, setting larger thresholds for bidirectional cases to reduce the number of sub-blocks and motion vectors, thereby correlating the complexity of bidirectional affine inter-prediction with unidirectional affine inter-prediction, and rounding motion vectors to a precision of one-sixteenth accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If bidirectional affine inter-prediction is implemented with fine sub-block granularity, then prediction accuracy is improved, but computational complexity increases significantly
Solution Approach 1:
The patent applies different minimum sub-block size parameters for unidirectional and bidirectional affine inter-prediction. Specifically, it sets a first minimum sub-block size for unidirectional case and a second larger minimum sub-block size for bidirectional case, thereby controlling the granularity of sub-block partitioning based on prediction type to balance accuracy and complexity
Solution Approach 2:
The patent implements local quality by applying different sub-block size constraints to different prediction modes. Unidirectional affine inter-prediction allows finer sub-block partitioning while bidirectional affine inter-prediction uses coarser partitioning, optimizing the balance between prediction precision and computational load for each specific case
2Measurement precision
If the number of sub-blocks is increased to improve prediction precision, then motion compensation accuracy is improved, but bandwidth usage increases
Solution Approach 1:
The patent changes the parameter of minimum sub-block size based on prediction mode. By setting a larger second minimum sub-block size for bidirectional affine inter-prediction compared to the first minimum sub-block size for unidirectional case, it reduces the total number of sub-blocks and their associated motion vectors, thereby reducing bandwidth usage while maintaining adequate prediction accuracy
3Ease of manufacture
If sub-block size thresholds are set uniformly for both prediction types, then implementation simplicity is maintained, but efficiency is reduced due to excessive computation in bidirectional cases
Solution Approach 1:
The patent implements local quality by differentiating sub-block size constraints between unidirectional and bidirectional affine inter-prediction modes. This targeted differentiation optimizes processing efficiency for the computationally intensive bidirectional case while maintaining straightforward implementation through clear conditional logic
Data Source
AI summary
A video coding device selects a current block for sub-block based affine inter-prediction, and derives a sub-block size for a sub-block of the current block. The sub-block size includes a sub-block width and a sub-block height. The device determines a unidirectional width threshold (TwU), a bidirectional width threshold (TwB), a unidirectional height threshold (ThU), and a bidirectional height threshold (ThB), where a total value of TwB and ThB exceeds a total value of TwU and ThU. The device determines that the affine inter-prediction is unidirectional inter-prediction or bidirectional inter-prediction. Based on the determination, the device applies the TwU or the TwB to the sub-block width, and applies the ThU or the ThB to the sub-block height. The device then derives a motion vector for the sub-block based on motion vectors for the current block, the sub-block width, and the sub-block height.