Sub-block Motion Derivation for Merge Mode Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding systems face challenges in efficiently deriving motion vectors for sub-blocks and refining motion vectors in merge mode, which affects video compression and decoding efficiency.
Innovation Solution
The proposed solution involves sub-block motion derivation and motion vector refinement techniques. Specifically, it identifies a collocated picture and selects a candidate neighboring CU based on temporal differences between reference pictures. It then derives motion vectors for sub-blocks using temporal scaling and replaces invalid motion vectors with those from valid collocated blocks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If motion vectors are derived for each sub-block using temporal scaling, then motion vector prediction accuracy is improved, but computational complexity increases
Solution Approach 1:
The current picture is divided into multiple sub-blocks, and motion vectors are derived separately for each sub-block using temporal scaling. This segmentation allows for more precise motion compensation by capturing local motion variations within different regions of the picture, thereby improving prediction accuracy while managing computational complexity through localized processing.
Solution Approach 2:
Different motion derivation methods are applied to different sub-blocks based on their local characteristics. By performing temporal scaling and motion vector derivation at the sub-block level rather than uniformly across the entire picture, the system adapts to local motion patterns, improving prediction accuracy in regions with varying motion while avoiding unnecessary computational overhead in uniform regions.
2Measurement precision
If multiple neighboring CUs are evaluated to select the best candidate, then motion vector accuracy is improved, but processing time increases
Solution Approach 1:
Instead of evaluating all possible neighboring CUs exhaustively, the method selects a limited set of candidate neighboring CUs based on spatial proximity and reference picture availability. This partial evaluation approach achieves sufficient motion vector accuracy by focusing on the most promising candidates, thereby reducing processing time while maintaining prediction quality.
Solution Approach 2:
The system pre-identifies candidate neighboring CUs based on their spatial relationship to the current CU and the availability of their reference pictures before performing detailed motion vector derivation. This preliminary selection filters out unlikely candidates early, reducing the number of expensive temporal scaling operations needed while ensuring the best candidates are evaluated.
3Measurement precision
If temporal scaling is applied to reference MV, then motion vector accuracy across different time points is improved, but computational overhead increases
Solution Approach 1:
The temporal scaling factor is computed based on the picture order count (POC) difference between the collocated picture and the reference picture. By dynamically adjusting the scaling parameter according to temporal distance, the method accurately compensates for motion changes over time. The scaling operation is applied selectively only when temporal differences exist, optimizing the balance between accuracy and computational overhead.
4Reliability
If invalid collocated blocks are replaced with valid ones, then reliability of motion vector derivation is improved, but additional processing steps are required
Solution Approach 1:
The system performs validity checks on collocated blocks before using them for motion vector derivation. By anticipating potential invalidity issues (such as out-of-bound blocks or blocks with unavailable reference pictures), the method prepares alternative valid collocated blocks in advance. This cushioning approach ensures reliable motion vector derivation without interrupting the decoding process, as replacements are already identified when needed.
Data Source
AI summary
Systems, methods, and instrumentalities for sub-block motion derivation and motion vector refinement for merge mode may be disclosed herein. Video data may be coded (e.g., encoded and/or decoded). A collocated picture for a current slice of the video data may be identified. The current slice may include one or more coding units (CUs). One or more neighboring CUs may be identified for a current CU. A neighboring CU (e.g., each neighboring CU) may correspond to a reference picture. A (e.g., one) neighboring CU may be selected to be a candidate neighboring CU based on the reference pictures and the collocated picture. A motion vector (MV) (e.g., collocated MV) may be identified from the collocated picture based on an MV (e.g., a reference MV) of the candidate neighboring CU. The current CU may be coded (e.g., encoded and/or decoded) using the collocated MV.