Block Level Motion Prediction for Video Coding Complexity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current sub-block based prediction methods in video coding face challenges such as high computational complexity, increased memory bandwidth, and undesirable division operations, which hinder coding efficiency and hardware implementation.
Innovation Solution
Applying motion prediction on a block level instead of a sub-block level, using affine motion compensation and combining spatial and temporal motion information to derive motion candidates, and simplifying motion vector prediction processes to reduce complexity and improve coding performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sub-block based prediction is used, then motion prediction accuracy is improved, but computational complexity and memory bandwidth increase
Solution Approach 1:
The patent merges the motion prediction process from sub-block level to block level, combining multiple sub-block predictions into a single block-level prediction. This reduces the number of separate prediction operations while maintaining overall prediction accuracy, thereby lowering computational complexity and memory bandwidth requirements.
Solution Approach 2:
The patent applies motion prediction at a coarser block level rather than the finer sub-block level, using partial prediction action that is sufficient for coding efficiency without the excessive computational overhead of sub-block level processing. This partial action approach achieves adequate prediction accuracy while significantly reducing complexity.
2Measurement precision
If sub-block based prediction is used, then motion prediction accuracy is improved, but memory bandwidth increases
Solution Approach 1:
The patent merges multiple sub-block motion predictions into a single block-level prediction, reducing the total amount of data that needs to be fetched and processed from memory. This consolidation decreases memory bandwidth requirements while preserving the essential motion information needed for accurate prediction.
Solution Approach 2:
The patent uses partial prediction at the block level that provides sufficient accuracy without the excessive memory access requirements of sub-block level prediction. This approach achieves the necessary prediction quality with significantly reduced memory bandwidth consumption.
3Measurement precision
If sub-block based prediction is used, then motion prediction accuracy is improved, but division operations increase
Solution Approach 1:
The patent merges the prediction process to eliminate the need for multiple division operations that would be required for sub-block level processing. By operating at the block level, the patent reduces the number of division operations while maintaining prediction accuracy through the unified motion model.
Solution Approach 2:
The patent uses partial prediction at the block level that avoids the excessive division operations required for sub-block prediction. This partial action approach achieves adequate accuracy without the computational burden of multiple division operations.
Data Source
AI summary
A method of video processing includes: deriving, from a set of neighboring blocks, a motion prediction of a current block which can be split into multiple sub-blocks; and performing a conversion between the current block and a bitstream representation of the current block using the derived motion prediction, the motion prediction being applied to the current block on a basis of a block level instead of on a sub-block level.