Bi-Prediction Mode Selection for Lower-Complexity Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding and decoding technologies face complexity issues due to the use of decoder side motion vector refinement (DMVR) and bi-predicted local illumination compensation (LIC) modes, which require multiple search rounds and iterations, increasing computational demands.
Innovation Solution
Implement techniques to determine and apply either DMVR mode or LIC mode for video blocks based on a merge candidate list, reducing the complexity by selecting only one mode for video encoding and decoding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If both DMVR mode and LIC mode are applied to video blocks, then the prediction accuracy and video quality are improved, but the computational complexity and resource utilization increase significantly
Solution Approach 1:
The patent changes the parameter of mode selection by introducing conditional logic that switches between DMVR and LIC modes based on block characteristics. Instead of always applying both modes, the system dynamically selects which mode to apply, thereby maintaining prediction accuracy where needed while reducing computational complexity in other cases.
Solution Approach 2:
The patent applies the principle of partial action by selectively applying either DMVR or LIC mode to only certain video blocks that meet specific criteria, rather than applying both modes to all blocks. This partial application reduces the overall computational burden while maintaining quality where the additional processing is most beneficial.
2Measurement precision
If multiple search rounds and iterations are performed in DMVR and LIC modes, then the motion vector refinement and illumination compensation accuracy are improved, but the encoding and decoding time increase
Solution Approach 1:
The patent changes the parameter of processing iterations by implementing conditional logic that limits the number of search rounds and iterations based on block characteristics. The system performs multiple iterations only when necessary to achieve the desired accuracy, while using fewer or no iterations for blocks where simple processing suffices, thereby reducing overall encoding and decoding time.
3Measurement precision
If DMVR mode is applied to bi-predicted blocks, then the motion compensation accuracy is improved, but the computational demands and resource utilization increase
Solution Approach 1:
The patent applies local quality by making the application of DMVR mode spatially selective - it is applied only to specific video blocks that meet certain criteria rather than uniformly to all blocks. This localized application ensures high motion compensation accuracy where needed while reducing computational demands in regions where simpler processing is sufficient.
Data Source
AI summary
A video encoder and video decoder are configured to code a block of video data using a merge candidate list that includes one or more bi-predicted merge candidates, determine to apply only one of a decoder side motion vector refinement (DMVR) mode or a local illumination compensation (LIC) mode to the block of video data, and apply one of the DMVR mode or the LIC mode to the block of video data based on the determination.