Affine Motion Model Candidate List for Inter Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image compression technologies face challenges in efficiently encoding and decoding high-resolution images, particularly in accurately predicting motion information for inter prediction.
Innovation Solution
The method involves using an affine model of an affine block in a non-affine mode to derive candidate motion information, which is then used to configure a merge candidate list for the current block, improving the accuracy and reliability of inter prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional inter prediction methods are used, then encoding complexity is reduced, but prediction accuracy deteriorates for high-resolution images
Solution Approach 1:
The patent changes the parameter of motion model selection by using affine motion models for blocks at specific positions (e.g., blocks at corners or edges of a picture) while using traditional non-affine models for other blocks. This selective parameter change improves prediction accuracy for regions where affine motion is more applicable without universally increasing encoding complexity across all blocks.
Solution Approach 2:
The patent segments the picture into different block types based on position characteristics. Blocks at specific positions (e.g., corner blocks, edge blocks) are identified and processed with affine motion models, while other blocks use traditional models. This segmentation allows the system to apply complex processing only where needed, improving overall prediction accuracy while controlling encoding complexity.
2Measurement precision
If affine motion models are used for all blocks, then prediction accuracy improves, but encoding complexity increases
Solution Approach 1:
The patent applies different motion model qualities to different spatial locations. Affine motion models (higher quality) are applied specifically to blocks at certain positions where motion complexity is higher, while traditional models (lower quality) are used for other blocks. This local quality approach ensures prediction accuracy where needed while avoiding unnecessary complexity elsewhere.
Solution Approach 2:
Instead of applying affine motion models to all blocks (excessive action), the patent applies them only to specific blocks at defined positions (partial action). This partial application provides sufficient prediction accuracy improvement for critical regions while avoiding the excessive encoding complexity that would result from universal application.
3Reliability
If more merge candidates are derived from affine blocks, then inter prediction reliability improves, but candidate list configuration complexity increases
Solution Approach 1:
The patent performs preliminary identification of affine blocks at specific positions before constructing the merge candidate list. By pre-identifying which blocks are affine and their positions, the system can efficiently derive motion candidates without complex real-time analysis during candidate list construction, thus improving reliability while controlling configuration complexity.
Solution Approach 2:
The patent copies motion information from identified affine blocks at specific positions to create merge candidates for the current block. This copying approach reliably transfers proven motion patterns from spatially related blocks while avoiding the need for complex motion estimation, thereby improving prediction reliability with minimal additional complexity.
Data Source
AI summary
An image decoding/encoding method and device according to the present disclosure may configure a merge candidate list of a current block, derive motion information of the current block on the basis of the merge candidate list and a merge index, and perform inter prediction for the current block on the basis of the motion information of the current block. Here, the merge candidate list may include a candidate derived using an affine motion model of an affine block coded by affine prediction.