Inter-Prediction Candidate Reordering for Bidirectional Motion Costing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image compression technologies face inefficiencies in reordering candidate lists for inter prediction, leading to suboptimal encoding and decoding processes.
Innovation Solution
An image decoding method and apparatus that reorder candidates in a candidate list based on a predetermined cost, using template matching-based or bilateral matching-based calculation methods, and selectively apply these methods based on block size, bidirectional prediction conditions, and position information to improve encoding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If template matching-based calculation method is used for cost calculation, then calculation speed is improved, but prediction accuracy deteriorates for bidirectional prediction cases
Solution Approach 1:
The patent implements dynamic selection between template matching-based and bilateral matching-based calculation methods based on prediction direction. For unidirectional prediction, template matching is used for fast calculation. For bidirectional prediction, bilateral matching is used for accurate cost calculation. This dynamic adaptation resolves the contradiction between calculation speed and prediction accuracy by selecting the appropriate method for each specific case.
Solution Approach 2:
The patent changes the calculation method parameter based on prediction direction conditions. When bidirectional prediction is detected, the system switches from template matching to bilateral matching calculation mode. This parameter change allows the system to maintain both speed (for unidirectional cases) and accuracy (for bidirectional cases) by adapting the calculation approach to the specific prediction scenario.
2Measurement precision
If bilateral matching-based calculation method is used for cost calculation, then prediction accuracy is improved, but calculation complexity and memory bandwidth requirements increase
Solution Approach 1:
The system dynamically selects the calculation method based on prediction direction conditions. Bilateral matching is applied only when bidirectional prediction is required, avoiding unnecessary complexity in unidirectional prediction cases. This dynamic approach ensures high accuracy when needed while minimizing overall calculation complexity.
Solution Approach 2:
The patent applies different calculation qualities to different prediction scenarios. Unidirectional prediction uses simple template matching (low complexity), while bidirectional prediction uses accurate bilateral matching (high complexity). This local quality differentiation ensures that computational resources are allocated efficiently based on the specific requirements of each prediction case.
3Productivity
If candidate list reordering is performed based on calculated cost, then encoding efficiency is improved, but processing time increases
Solution Approach 1:
The system dynamically determines whether candidate list reordering is necessary based on prediction direction. Reordering is performed only when bidirectional prediction is used and cost calculation is required. For unidirectional prediction, the default candidate order is maintained without reordering operations. This dynamic approach improves encoding efficiency when beneficial while minimizing processing time overhead.
Solution Approach 2:
The patent changes the processing flow parameter based on prediction direction conditions. When bidirectional prediction is detected, the system activates candidate list reordering based on calculated costs. When unidirectional prediction is used, the system skips reordering and uses default candidates directly. This parameter change optimizes the balance between encoding efficiency and processing time.
Data Source
AI summary
An image decoding/encoding method and apparatus, according to the present disclosure, may generate a candidate list for a current block, reorder, on the basis of a certain cost, one or more candidates included in the candidate list, and derive motion information regarding the current block on the basis of the candidate list, so as to perform inter-prediction. In this regard, the cost may be calculated for each of a plurality of candidates included in the candidate list on the basis of at least one of a template matching-based calculation method, a two-way matching-based calculation method, or a motion information-based calculation method.