Motion-Based Candidate Lists for Efficient Inter-Prediction Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video codecs face inefficiencies in generating candidate motion vector lists for inter-prediction, leading to increased bit rates for signaling motion vectors due to predetermined ordering based on block location rather than motion information.
Innovation Solution
The proposed solution involves ordering candidate motion vector lists based on motion information such as quantization parameters, temporal distance, and coding type, prioritizing more likely candidates to reduce the number of bits needed for signaling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If candidate motion vector lists are ordered based on block location (predetermined order), then the construction process is simple and fast, but more bits are required for signaling motion vectors
Solution Approach 1:
The patent changes the ordering parameter from spatial location to motion information characteristics. By sorting candidates based on motion vector magnitude, reference picture type, and temporal distance rather than their position in the block, the system achieves more efficient bit rate usage while maintaining manageable complexity through defined sorting criteria.
Solution Approach 2:
The candidate list ordering becomes dynamic rather than static. Instead of a fixed predetermined order based on block location, the list is dynamically reordered based on motion information characteristics that vary with each coding unit, allowing the most promising candidates to be positioned earlier in the list for more efficient signaling.
2Productivity
If candidate motion vector lists are ordered based on motion information, then coding efficiency is improved, but the construction process becomes more complex
Solution Approach 1:
The patent employs parameter changes by sorting candidates according to motion information parameters such as motion vector magnitude, reference picture type (L0/L1), and temporal distance. This reordering based on motion characteristics improves coding efficiency by placing more likely candidates earlier in the list, while the complexity is controlled through well-defined sorting criteria.
Solution Approach 2:
The system performs preliminary sorting of candidate motion vectors based on motion information before the actual encoding decision is made. By pre-ordering the candidate list according to motion characteristics, the encoder prepares an optimized sequence that speeds up the subsequent encoding process and improves overall coding efficiency.
3Loss of information
If more candidates are placed earlier in the list based on motion likelihood, then fewer bits are needed for signaling, but more computational resources are required for analysis
Solution Approach 1:
The patent uses parameter changes by ordering candidates based on motion information parameters that can be computed with relatively low complexity, such as motion vector magnitude, reference picture type, and temporal distance. This approach achieves efficient bit rate reduction while avoiding the need for computationally intensive analysis, as these parameters are already available or easily derived from the motion compensation process.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Embodiments include techniques for generation of candidate motion vector lists for use in inter-prediction. For example, according to some embodiments, a candidate motion vector list is generated to have an order that is based on motion information of each of the candidate motion vectors. The order can be applied as the list is generated, as a resorting after the list is partially generated, or after the list is generated.