Triangle Merge Mode Candidate List Derivation Simplification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The candidate list derivation for the triangle Merge mode in video coding is complex and involves heavy data dependency, making it difficult to speed up the process effectively.
Innovation Solution
A method is proposed to generate a first candidate list and then a second candidate list for the triangle Merge mode, reusing the normal Merge candidate list and modifying bi-prediction candidates to uni-prediction candidates, with specific pruning processes to reduce redundancy and simplify the derivation process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the normal Merge candidate list derivation process is used for triangle Merge mode, then the coding accuracy is maintained, but the computational complexity and data dependency increase significantly
Solution Approach 1:
The candidate list for triangle Merge mode is segmented into two separate lists: a first candidate list containing spatial candidates and a second candidate list containing temporal candidates. This segmentation allows each list to be derived independently with reduced data dependency, while together they provide comprehensive motion prediction candidates maintaining coding accuracy.
Solution Approach 2:
The invention transitions from a single unified candidate list to a two-dimensional candidate list structure (spatial and temporal dimensions). The first candidate list derives spatial motion candidates from neighboring blocks, while the second candidate list derives temporal motion candidates from co-located blocks in reference pictures, adding a temporal dimension to the candidate selection process.
2Measurement precision
If the complex pruning processes are applied to remove redundancy in triangle Merge mode, then the candidate quality is improved, but the processing time and computational load increase
Solution Approach 1:
The invention performs preliminary organization of motion candidates by dividing them into spatial and temporal categories before final selection. The first candidate list is prepared with spatial candidates and the second with temporal candidates, allowing the decoder to efficiently process and select from pre-organized lists without requiring complex post-processing pruning operations.
3Measurement precision
If bi-prediction candidates are used in triangle Merge mode, then the prediction accuracy is improved, but the data dependency and processing complexity increase
Solution Approach 1:
The bi-prediction capability is segmented into separate spatial and temporal components. The first candidate list handles spatial bi-prediction candidates from neighboring blocks, while the second candidate list handles temporal bi-prediction candidates from reference pictures. This segmentation reduces data dependency by allowing independent derivation of each list.
Data Source
AI summary
A method and apparatus of Inter prediction for video coding using a target Merge mode comprising a triangle Merge mode are disclosed. According to this method, a current block is partitioned into a first region and a second region using the target partition. A candidate list is determined. Two candidates are determined for the two regions by deriving two target candidates from the candidate list. When a selected target candidate is a uni-prediction candidate, the selected target candidate is used as one candidate; and when the selected target candidate is a bi-prediction candidate, motion information associated of the selected target candidate with List 0 or List 1 is used as one candidate. The current block or current motion information associated with the current block is encoded or decoded according to the candidate list. The candidate list is used to derive the first candidate and the second candidate.