Temporal Motion Data Derivation Within LCU Boundaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards, such as HEVC, face challenges in efficiently deriving temporal motion data candidates due to misalignment issues that result in high memory bandwidth consumption and cycle overhead during the derivation of temporal motion data candidates, particularly when fetching motion data across LCU boundaries.
Innovation Solution
Constraining temporal motion data positions to be within the boundaries of the co-located LCU, thereby optimizing memory bandwidth usage and reducing cycle overhead by aligning the fetch area with the current LCU.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If temporal motion data positions are extended beyond co-located LCU boundaries to improve motion data availability, then motion compensation accuracy is improved, but memory bandwidth consumption and cycle overhead increase
Solution Approach 1:
The patent segments the motion data fetch area to be confined within the co-located LCU boundaries, dividing the reference picture into distinct LCU regions. This segmentation allows the encoder to fetch temporal motion data only from within the co-located LCU, avoiding cross-LCU boundary accesses that would increase memory bandwidth consumption while still providing sufficient motion data candidates for accurate motion compensation.
2Measurement precision
If temporal motion data positions are extended beyond co-located LCU boundaries to improve motion data availability, then motion compensation accuracy is improved, but processing time and cycle overhead increase
Solution Approach 1:
The patent segments the motion data fetch area to be confined within the co-located LCU boundaries, dividing the reference picture into distinct LCU regions. This segmentation allows the encoder to fetch temporal motion data only from within the co-located LCU, avoiding cross-LCU boundary accesses that would increase memory bandwidth consumption and processing time, while still providing sufficient motion data candidates for accurate motion compensation.
3Measurement precision
If temporal motion data positions are extended beyond co-located LCU boundaries to improve motion data availability, then motion compensation accuracy is improved, but device complexity increases due to additional hardware requirements
Solution Approach 1:
The patent segments the motion data fetch area to be confined within the co-located LCU boundaries, eliminating the need for additional hardware structures like line buffers that would be required to handle cross-LCU boundary accesses. This segmentation approach maintains motion compensation accuracy while reducing device complexity by utilizing the existing LCU-based memory organization.
Data Source
AI summary
A method for derivation of a temporal motion data (TMD) candidate for a prediction unit (PU) in video encoding or video decoding is provided. The derived TMD candidate is for inclusion in an inter-prediction candidate list for the PU. The method includes determining a primary TMD position relative to a co-located PU in a co-located largest coding unit (LCU), wherein the co-located PU is a block in a reference picture having a same size, shape, and coordinates as the PU, and selecting at least some motion data of a secondary TMD position as the TMD candidate when the primary TMD position is in a bottom neighboring LCU or in a bottom right neighboring LCU of the co-located LCU, wherein the secondary TMD position is determined relative to the co-located PU.