Bi-Prediction Weight Adaptation for Lower-Complexity Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding systems face high computational complexity due to exhaustive searches for optimal motion vectors and weights in generalized bi-prediction, especially in conditions with rapid illuminance changes, leading to inefficiencies in coding efficiency and increased processing time.
Innovation Solution
Adaptive techniques are employed to reduce the number of generalized bi-prediction weights based on temporal layers, quality of reference pictures, and similarity of prediction signals, combined with early termination and weight reuse strategies to simplify the encoding process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If exhaustive search for optimal motion vectors and weights is performed in generalized bi-prediction, then coding precision is improved, but device complexity increases
Solution Approach 1:
The patent segments the exhaustive search process into multiple stages: first searching for optimal motion vectors, then searching for optimal weights only in regions where motion vectors are similar. This segmentation reduces the overall search space and computational complexity while maintaining coding precision in critical areas.
Solution Approach 2:
The patent applies different levels of search intensity to different regions of the video block. In regions with similar motion vectors, a reduced weight search is performed, while in regions with significant motion variation, full search is maintained. This local quality approach optimizes the balance between coding precision and computational complexity.
2Manufacturing precision
If exhaustive search for optimal motion vectors and weights is performed in generalized bi-prediction, then manufacturing precision is improved, but productivity decreases
Solution Approach 1:
The patent divides the computational workload into segments: motion vector search is performed first, followed by a conditional weight search only in blocks where motion vectors are similar. This segmentation reduces total processing time and improves productivity while maintaining coding precision where it matters most.
Solution Approach 2:
The patent performs partial action by conducting a reduced weight search only in blocks with similar motion vectors, rather than performing exhaustive search in all blocks. This partial approach maintains sufficient coding precision while significantly improving processing throughput and productivity.
3Device complexity
If adaptive techniques reduce the number of generalized bi-prediction weights based on temporal layers and similarity, then device complexity is reduced, but manufacturing precision may deteriorate
Solution Approach 1:
The patent applies adaptive weight reduction selectively based on local characteristics: blocks with similar motion vectors use reduced weight sets, while blocks with significant motion variation maintain full weight search. This local quality approach ensures coding precision is maintained in critical regions while reducing complexity elsewhere.
Solution Approach 2:
The patent implements dynamic adaptation of the weight search space based on motion vector similarity and temporal layer information. The weight search range is dynamically adjusted according to block characteristics, allowing the system to maintain high coding precision when needed while reducing complexity in appropriate scenarios.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Exemplary embodiments include systems and methods for coding a video comprising a plurality of pictures including a current picture, a first reference picture, and a second reference picture, where each picture includes a plurality of blocks. In one method, for at least a current block in the current picture, a number of available bi-prediction weights is determined based at least in part on a temporal layer and/or a quantization parameter of the current picture. From among available bi-prediction weights a pair of weights are identified. Using the identified weights, the current block is then predicted as a weighted sum of a first reference block in the first reference picture and a second reference block in the second reference picture. Encoding techniques are also described for efficient searching and selection of a pair of bi-prediction weights to use for prediction of a block.