Video Block Coding with Split-Line Affine Motion Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current geometric partitioning mode (GPM) in video coding is inadequate for predicting blocks with complex motions such as zoom, rotation, and perspective due to its reliance on a translation motion model.
Innovation Solution
The method involves determining block units with split lines and using affine candidates to predict blocks by combining first and second predicted blocks based on affine motion information, incorporating affine merge and motion vector prediction candidates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If translation motion model is used in GPM, then device complexity is reduced, but manufacturing precision deteriorates because it cannot accurately predict blocks with complex motions such as zoom, rotation, and perspective
Solution Approach 1:
The patent changes the motion model parameters from simple translation to affine transformation parameters. The affine motion model uses parameters including rotation angle, zoom scale, and translation components, allowing it to accurately represent complex motions while maintaining computational efficiency through parameterized modeling.
Solution Approach 2:
The patent introduces dynamic motion compensation by applying affine transformation that can adapt to different types of motion (rotation, zoom, translation) within the same block. This dynamic approach allows the prediction to adjust to the actual motion characteristics rather than assuming uniform translation.
2Manufacturing precision
If affine motion model is applied to predict blocks with diverse motions, then prediction accuracy is improved, but device complexity increases due to additional calculations and parameters
Solution Approach 1:
The patent segments the block prediction process by dividing the block into sub-blocks and applying different affine motion parameters to each sub-block. This segmentation allows the complex affine transformation to be broken down into manageable pieces, reducing computational complexity while maintaining accuracy.
Solution Approach 2:
The patent applies affine motion model selectively rather than universally. By using partial action (applying affine only where needed based on motion complexity) and excessive action (using sufficient but not overly complex parameters), the system achieves good prediction accuracy without unnecessary computational burden.
3Loss of information
If only two merge candidates are selected in GPM, then bitstream overhead is reduced, but prediction accuracy deteriorates for blocks with complex motion patterns
Solution Approach 1:
The patent adds another dimension to the motion representation by using affine transformation parameters (rotation angle, zoom scale, translation) instead of simple translation vectors. This dimensional enrichment allows two candidates to convey more motion information, maintaining low bitstream overhead while improving prediction accuracy for complex motions.
Data Source
AI summary
A method of decoding video data by an electronic device is provided. The electronic device determines, according to the video data, a block unit from an image frame and a split line of the block unit. The electronic device determines first motion information and second motion information according to the video data. The electronic device determines a first predicted block of the block unit based on the first motion information by an affine model and determines a second predicted block of the block unit based on the second motion information. The block unit is predicted by combining the first predicted block and the second predicted block based on the split line.