Affine Merge Mode Motion Vector Candidate Derivation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards face challenges in efficiently deriving motion vector candidates for affine motion prediction mode, leading to suboptimal compression efficiency and video quality.
Innovation Solution
The proposed method improves motion vector candidate derivation by obtaining restricted areas not adjacent to the current coding unit, projecting scanning positions to adjacent pixel positions, and deriving motion vector candidates and control point motion vectors from non-adjacent coding units.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If motion vector candidates are derived only from adjacent CUs, then the derivation process is simple and fast, but the accuracy and compression efficiency are suboptimal
Solution Approach 1:
The patent segments the candidate derivation process into two distinct parts: adjacent CU candidates (spatially close) and non-adjacent CU candidates (projected from scanning positions). This segmentation allows the system to expand the search space to improve accuracy while maintaining a structured, manageable derivation process by treating different spatial regions separately.
Solution Approach 2:
The patent introduces a new dimension to the candidate search by projecting scanning positions from the current picture onto reference pictures. Instead of only searching in the spatial domain of adjacent blocks, the system now searches in the temporal dimension by mapping positions across multiple reference pictures, thereby expanding the candidate pool without proportionally increasing complexity.
2Productivity
If more CUs are used to derive MV candidates, then compression efficiency improves, but computational complexity increases
Solution Approach 1:
The patent performs preliminary action by pre-calculating and storing the mapping relationships between scanning positions in the current picture and corresponding positions in reference pictures. This pre-computation allows the decoder to quickly derive non-adjacent candidates during decoding without performing complex real-time calculations, thereby improving compression efficiency while controlling computational complexity.
Solution Approach 2:
The patent uses copying by reusing motion vector information from non-adjacent CUs that are projected from scanning positions. Instead of independently calculating motion vectors for each candidate, the system copies and adapts motion vector data from reference positions, significantly reducing computational complexity while maintaining improved compression efficiency through a larger candidate pool.
3Adaptability or versatility
If scanning positions are projected to non-adjacent pixel positions, then more diverse candidates are obtained, but the derivation process becomes more complex
Solution Approach 1:
The patent applies parameter changes by systematically varying the scanning positions and their corresponding projection parameters (reference picture index, offset values) to generate diverse non-adjacent candidates. By controlling and parameterizing the projection process, the system achieves high candidate diversity while keeping the derivation complexity manageable through structured parameter management rather than uncontrolled search.
Data Source
AI summary
Methods for video decoding and encoding, apparatuses and non-transitory storage media are provided. In one decoding method, the decoder obtains a restricted area that is not adjacent to a current coding unit (CU) according to a value associated with the restricted area. Additionally, the decoder obtains one or more motion vector (MV) candidates from a plurality of non-adjacent CUs to the current CU based on the restricted area. Furthermore, the decoder obtains one or more control point motion vectors (CPMVs) for the current CU based on the one or more MV candidates.