Displacement Vector Prediction Using Multi-Mode Candidate Construction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video encoding and decoding standards, such as VVC and AVS3, face inefficiencies in displacement vector prediction, which hinders the improvement of encoding performance.
Innovation Solution
A method and apparatus for displacement vector prediction in video encoding and decoding that constructs candidate displacement vectors from multiple prediction modes and derives a predicted displacement vector, enhancing encoding performance by utilizing information from both same and different prediction modes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If a history-based block vector prediction (HBVP) list is constructed based on a block vector of a historical block, then the prediction process is simplified, but the encoding performance is insufficient
Solution Approach 1:
The patent merges HBVP list construction with intra-block copy (IBC) prediction by using IBC reference blocks to generate candidate block vectors. This combines the simplicity of history-based prediction with the accuracy of intra-block copy, resolving the contradiction between process simplicity and encoding performance.
Solution Approach 2:
The patent makes the HBVP list construction method universal by applying it to multiple prediction modes including IBC and inter prediction. The same mechanism of using historical block vectors and IBC reference blocks serves multiple prediction purposes, improving encoding performance across different modes while maintaining a unified simple process.
2Measurement precision
If a class-based block vector prediction (CBVP) list is derived from the HBVP list, then the prediction accuracy is improved, but the computational complexity increases
Solution Approach 1:
The patent applies local quality by differentiating CBVP list construction based on the current block's prediction mode. For IBC mode, IBC reference blocks are used; for inter mode, different reference blocks are used. This localized approach improves prediction accuracy for each specific case while avoiding unnecessary complexity in other modes.
Solution Approach 2:
The patent performs preliminary action by pre-construction of the HBVP list using IBC reference blocks before deriving the CBVP list. This preliminary construction of high-quality candidate vectors simplifies the subsequent CBVP derivation process, improving accuracy while reducing the computational complexity of the overall two-stage process.
3Reliability
If displacement vectors from multiple prediction modes are combined, then the likelihood of selecting optimal candidate vectors increases, but the bit overhead increases
Solution Approach 1:
The patent applies partial action by selectively combining displacement vectors from multiple prediction modes only when beneficial. The method constructs candidate lists with vectors from both IBC and inter prediction modes, but uses mode-specific logic to include only relevant vectors, achieving optimal candidate selection while controlling bit overhead through selective inclusion rather than exhaustive combination.
Data Source
AI summary
This application provides a displacement vector prediction method and apparatus in video encoding/decoding and a device and relates to the field of video codec technologies. In this application, candidate displacement vectors are constructed by combining displacement vectors related to a plurality of prediction modes, and then a predicted displacement vector of a current unit is derived from the candidate displacement vectors. Information about an encoded/decoded unit having a same prediction mode as the current unit can be used and information about an encoded/decoded unit having different prediction modes from the current unit is also allowed to be combined to derive the predicted displacement vector of the current unit, so that the candidate displacement vector has a higher possibility to be selected as the predicted displacement vector, thereby improving the encoding performance of the displacement vector.