Uni-predictive Prediction Units Reduce Memory Bandwidth in HEVC
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High memory bandwidth requirements for motion compensation in video coding, particularly in HEVC, pose a bottleneck, especially for smaller prediction unit sizes and Ultra High Definition video, necessitating a reduction in memory bandwidth usage.
Innovation Solution
Restricting prediction of prediction units in bi-predicted slices to uni-prediction when they reach a predetermined size, thereby reducing memory bandwidth needs by forcing PUs of specified restricted sizes to be coded only in uni-predictive mode.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If bi-prediction is used for small prediction units in HEVC, then prediction accuracy is improved, but memory bandwidth requirements increase significantly
Solution Approach 1:
The patent applies local quality by differentiating the prediction mode based on prediction unit size. Small PUs (4×4, 8×4, 4×8) are restricted to uni-prediction to reduce memory bandwidth, while larger PUs can use bi-prediction for higher accuracy. This localized approach optimizes the trade-off between prediction quality and memory usage for different spatial regions of the video data.
Solution Approach 2:
The patent changes the prediction mode parameter based on PU size. By modifying the prediction mode from bi-prediction to uni-prediction for small PUs, the system reduces memory bandwidth requirements while maintaining acceptable prediction quality. This parameter change is driven by the relationship between PU size and memory access patterns.
2Productivity
If HEVC uses larger block sizes and 8-tap filters, then coding efficiency is improved, but memory bandwidth requirements for motion compensation increase
Solution Approach 1:
The patent applies local quality by restricting bi-prediction to larger PUs only. For small PUs, uni-prediction is used which reduces memory bandwidth consumption. This creates a localized optimization where the prediction method adapts to the spatial scale of the processing unit, balancing coding efficiency with memory bandwidth constraints.
Solution Approach 2:
The patent applies partial action by not enabling bi-prediction for all PUs, but only for larger ones where the memory bandwidth benefit is less critical. This partial application of bi-prediction maintains coding efficiency for large blocks while reducing overall memory bandwidth requirements by excluding small PUs from bi-prediction.
3Quantity of substance
If inter-prediction of 4×4 prediction blocks is disabled, then memory bandwidth requirements are reduced, but prediction flexibility is limited
Solution Approach 1:
The patent applies local quality by allowing different prediction modes for different PU sizes. Small PUs (4×4, 8×4, 4×8) are restricted to uni-prediction to reduce memory bandwidth, while larger PUs retain bi-prediction flexibility. This localized differentiation maintains prediction flexibility where needed while reducing memory bandwidth where possible.
Solution Approach 2:
The patent applies dynamics by making the prediction mode adaptable based on PU size. The system dynamically selects between uni-prediction and bi-prediction based on the spatial dimensions of the prediction unit, providing flexibility for large PUs while optimizing for memory efficiency in small PUs.
Data Source
AI summary
Motion compensation requires a significant amount of memory bandwidth, especially for smaller prediction unit sizes. The worst case bandwidth requirements can occur when bi-predicted 4×8 or 8×4 PUs are used. To reduce the memory bandwidth requirements for such smaller PUs, methods are provided for restricting inter-coded PUs of small block sizes to be coded only in a uni-predictive mode, i.e., forward prediction or backward prediction. More specifically, PUs of specified restricted sizes in bi-predicted slices (B slices) are forced to be uni-predicted.