Motion Vector Offset Refinement for Merge-Based Video Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing demand for high-definition video services has led to a significant increase in data volume, and existing video compression standards like HEVC are showing limitations in performance.
Innovation Solution
A method of refining a motion vector using an offset vector in video signal encoding and decoding, involving the generation of a merge candidate list, determining a merge candidate, deriving an offset vector, and applying it to a motion vector, with adaptive determination of the offset vector's magnitude and direction based on index information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If HEVC compression standard is used to reduce data volume, then compression performance is improved, but the compression performance gradually reveals limitations as display panels get bigger and video quality requirements increase
Solution Approach 1:
The patent changes the parameters of motion vector representation by introducing offset vectors that can be derived from merge candidates. Instead of using fixed motion vector precision, the system adapts the motion vector parameters based on the offset vector magnitude and direction, allowing for more precise motion representation in high-definition video without increasing data volume significantly
Solution Approach 2:
The patent makes the motion vector refinement process dynamic by adaptively determining offset vector magnitude and direction based on index information from merge candidates. The system dynamically adjusts motion vector precision and offset values according to the specific characteristics of each block and merge candidate, enabling optimized compression performance for varying video content
2Measurement precision
If motion vector precision is increased to improve inter prediction accuracy, then prediction efficiency is improved, but the complexity of motion vector derivation and signaling increases
Solution Approach 1:
The patent introduces offset vectors as intermediary elements between merge candidates and final motion vectors. Instead of directly computing high-precision motion vectors, the system uses offset vectors as a mediator to refine merge candidate motion vectors. This intermediary approach simplifies the derivation process while achieving improved precision through the offset adjustment mechanism
Solution Approach 2:
The patent applies partial refinement by using offset vectors with limited precision (e.g., half-pel or quarter-pel accuracy) rather than full precision motion vectors. This partial action approach provides sufficient prediction accuracy for most cases while significantly reducing the complexity of motion vector derivation and signaling overhead
3Productivity
If offset vector magnitude and direction are adaptively determined based on index information, then inter prediction efficiency is improved, but the signaling overhead for indicating numerical range and motion magnitude candidates increases
Solution Approach 1:
The patent applies local quality by determining offset vector magnitude and direction adaptively based on index information specific to each block and merge candidate. Instead of using uniform high-precision motion vectors throughout, the system locally optimizes the offset vector parameters based on the specific characteristics of each block, achieving improved prediction efficiency where needed while maintaining lower precision where sufficient
Solution Approach 2:
The patent performs preliminary action by pre-defining a set of possible offset vector magnitudes and directions based on index information from merge candidates. The system prepares these candidate offset values in advance and selects from them during the encoding/decoding process, reducing the need for complex real-time calculations and minimizing signaling overhead while maintaining high prediction efficiency
Data Source
AI summary
A video decoding method according to the present disclosure includes the steps that: a merge candidate list for a current block is generated; a merge candidate for the current block is determined among merge candidates included in the merge candidate list; an offset vector for the current block is derived; and a motion vector for the current block is derived by adding the offset vector to a motion vector of the merge candidate.