Motion Vector Rounding for Video Decoder Memory Bandwidth Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies face challenges in reducing memory bandwidth requirements due to the use of sub-pixel motion vectors, which increase data transfer needs for interpolation, especially in scenarios where increasing memory bandwidth is not economical.
Innovation Solution
Rounding the horizontal and vertical components of motion vectors from sub-pixel accuracy to integer-pixel accuracy for specific prediction unit sizes, such as 4×8 or 8×4, reduces the data required for interpolation, thereby decreasing memory bandwidth demands.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sub-pixel motion vectors are used for interpolation, then prediction accuracy is improved, but memory bandwidth requirements increase
Solution Approach 1:
The patent applies different motion vector precision requirements to different prediction unit sizes. Specifically, 4×8 and 8×4 PUs use integer-pixel accuracy motion vectors, while other PU sizes use sub-pixel accuracy. This local differentiation allows the system to reduce memory bandwidth requirements for specific PU types without compromising prediction accuracy for other PU types where sub-pixel precision remains beneficial.
Solution Approach 2:
The patent changes the precision parameter of motion vectors from sub-pixel accuracy to integer-pixel accuracy for specific prediction unit sizes (4×8 and 8×4). This parameter change reduces the amount of data that needs to be transferred from memory during interpolation, thereby reducing memory bandwidth requirements while maintaining acceptable prediction quality for these specific PU sizes.
2Manufacturing precision
If sub-pixel motion vectors are used, then coding precision is improved, but data transfer requirements increase
Solution Approach 1:
The patent implements local quality by applying integer-pixel accuracy motion vectors specifically to 4×8 and 8×4 prediction units, while maintaining sub-pixel accuracy for other PU sizes. This targeted approach reduces data transfer requirements for specific PU types without sacrificing coding precision in other areas where sub-pixel precision provides measurable benefits.
Solution Approach 2:
The patent applies partial action by rounding motion vectors to integer-pixel accuracy only for specific prediction unit sizes (4×8 and 8×4) rather than uniformly for all PUs. This partial application of rounding reduces data transfer requirements in cases where sub-pixel precision is less critical, while preserving full precision where it provides the most value.
Data Source
AI summary
A video decoder determines, based at least in part on a size of a prediction unit (PU), whether to round either or both a horizontal or a vertical component of a motion vector of the PU from sub-pixel accuracy to integer-pixel accuracy. The video decoder generates, based at least in part on the motion vector, a predictive sample block for the PU and generates, based in part on the predictive sample block for the PU, a reconstructed sample block.