Video Merge-Mode Inter-Prediction with Combined Candidates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video compression technologies face challenges in efficiently encoding and decoding high-resolution and high-quality videos, particularly in constructing accurate merge candidate lists and performing inter-prediction, especially for 360-degree videos.
Innovation Solution
A method and apparatus for constructing a merge candidate list that includes spatial, temporal, and combined merge candidates, using weighted averages for motion information derivation, and employing an affine model for inter-prediction in units of sub-blocks, with consideration for reference region boundaries.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional inter-prediction methods are used, then the encoding/decoding process is simple, but the prediction accuracy is insufficient for high-resolution and high-quality videos
Solution Approach 1:
The patent divides the current block into multiple sub-blocks (e.g., four sub-blocks arranged in a 2x2 grid) and performs inter-prediction independently on each sub-block using sub-block motion vectors. This segmentation allows more flexible motion modeling that adapts to local motion characteristics, improving prediction accuracy while maintaining manageable complexity through standardized partitioning rules.
Solution Approach 2:
The patent introduces sub-block level prediction as an additional dimension of motion modeling, moving from block-level homogeneous motion to sub-block-level non-uniform motion. By deriving motion information at the sub-block dimension rather than only at the block dimension, the system captures spatial variations in motion more accurately, enhancing prediction quality for high-resolution videos.
2Measurement precision
If merge candidate list construction is simplified, then the processing time is reduced, but the accuracy of motion information derivation is insufficient
Solution Approach 1:
The patent performs preliminary construction of merge candidates using spatial and temporal neighboring blocks before final motion information derivation. By pre-identifying candidate blocks and preparing their motion vectors in advance, the system reduces the computational burden during final prediction while ensuring accurate motion information is derived from the most relevant neighboring blocks.
Solution Approach 2:
The patent introduces combined merge candidates as an intermediary mechanism that integrates information from multiple spatial and temporal neighboring blocks. This intermediary approach synthesizes motion information from various sources (left, right, above, below blocks) to produce more accurate motion vectors, balancing processing time with prediction accuracy through weighted averaging and selection criteria.
3Reliability
If block-based merge mode is used, then the implementation is simple, but it cannot accurately represent non-uniform motion in high-resolution videos
Solution Approach 1:
The patent segments each block into multiple sub-blocks and applies independent motion modeling to each sub-block. This allows the system to represent non-uniform motion patterns that vary across different regions of a block, improving reliability of motion representation while maintaining a systematic approach through standardized sub-block partitioning.
Solution Approach 2:
The patent enables dynamic adaptation between block-based and sub-block-based prediction modes. The system can switch between homogeneous block-level prediction and non-uniform sub-block-level prediction based on motion characteristics, allowing flexible representation of complex motion patterns while maintaining implementation simplicity through conditional logic.
Data Source
AI summary
A video encoding/decoding method is provided, which includes constructing a merge candidate list of a current block, deriving motion information of the current block from the merge candidate list, and performing inter-prediction of the current block using the motion information, where the merge candidate list includes at least one of a spatial merge candidate, a temporal merge candidate, or a combined merge candidate, and the combined merge candidate is derived by combining n merge candidates belonging to the merge candidate list. A video encoding/decoding apparatus is also provided.