Affine Motion Prediction Subblock Segmentation for Video Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video compression techniques, such as H.264/AVC and HEVC, face challenges in efficiently encoding and decoding video data due to increasing picture sizes, resolutions, and frame rates, necessitating improved encoding efficiency and image quality, particularly in inter-prediction methods.
Innovation Solution
A coding tool that employs affine motion prediction using control point motion vectors to generate and modify sample values for subblocks within a target block, compensating for various object motions, including translation, rotation, and zoom, by performing affine motion prediction on a subblock-by-subblock basis and adjusting sample values based on motion vector positions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional inter-prediction is used for video encoding, then encoding process is simple, but compression performance is insufficient for high resolution and high frame rate videos
Solution Approach 1:
The target block is divided into multiple subblocks, and affine motion prediction is performed independently for each subblock using control point motion vectors. This segmentation allows the system to handle complex motions in high resolution videos by treating each subblock separately, thereby improving compression performance without requiring a complete redesign of the encoding framework
Solution Approach 2:
The patent introduces affine motion models that can dynamically adapt to different motion patterns within a block. By using control point motion vectors and performing predictions on a subblock-by-subblock basis, the system can dynamically adjust to translation, rotation, and zoom motions, improving compression efficiency for diverse video content
2Measurement precision
If affine motion prediction is performed on the entire target block, then motion compensation accuracy is improved, but computational complexity increases significantly
Solution Approach 1:
The target block is divided into multiple subblocks, and affine motion prediction is performed independently for each subblock. This segmentation reduces the computational complexity by breaking down a large complex calculation into smaller, more manageable subcalculations, while still maintaining overall motion compensation accuracy
Solution Approach 2:
Different affine motion predictions are applied to different subblocks within the target block based on their local motion characteristics. This local quality approach ensures that each subblock receives the appropriate level of motion compensation accuracy, improving overall precision without uniformly increasing complexity across the entire block
3Adaptability or versatility
If control point motion vectors are used for subblock prediction, then various object motions are compensated, but the amount of data to be encoded increases
Solution Approach 1:
The control point motion vectors serve multiple purposes: they are used for affine motion prediction in all subblocks and can also represent the overall motion of the target block. This multi-functionality allows the system to compensate for various object motions without proportionally increasing the data volume, as the same control points serve multiple prediction needs
Data Source
AI summary
A video decoding apparatus is provided for decoding pictures using inter-prediction. The video decoding apparatus includes a predictor that generates affine prediction samples for respective subblocks in a target block in a current picture to be decoded by performing affine motion prediction on a subblock-by-subblock basis using control point motion vectors which respectively correspond to corners of the target block. The predictor also modifies sample values of the affine prediction samples by executing a first coding tool, the first coding tool compensating for motion according to sample positions in each of the subblocks using the control point motion vectors.