Adaptive Motion Coding for Multi-Hypothesis Video Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding techniques face challenges in efficiently encoding and decoding video data, particularly in reducing redundancy and improving prediction accuracy, especially when dealing with multiple hypothesis inter-prediction.
Innovation Solution
The proposed solution involves using multiple hypothesis inter-prediction techniques during video coding, specifically by generating and combining multiple prediction blocks for a current block, and signaling motion information using merge mode to reduce overhead and processing complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple hypothesis inter-prediction is used to improve prediction accuracy, then prediction accuracy is improved, but signaling overhead and processing complexity increase
Solution Approach 1:
The patent combines multiple prediction hypotheses (base hypothesis and additional hypothesis) into a unified multi-hypothesis prediction block. Motion information from multiple neighboring blocks is merged and adapted to form candidate motion information for both base and additional hypotheses, reducing redundant signaling. The decoder combines prediction blocks from multiple hypotheses to form the final prediction, achieving improved accuracy while managing complexity through systematic integration of multiple sources.
Solution Approach 2:
The patent performs preliminary adaptation of motion information from neighboring blocks before generating prediction blocks. Merge candidate lists are constructed in advance with adapted motion information, and the decoder prepares multiple prediction hypotheses beforehand. This preliminary processing organizes the complexity upfront, allowing the actual decoding to use pre-prepared motion information and prediction blocks, thereby reducing real-time processing burden while maintaining multiple hypotheses for improved accuracy.
2Measurement precision
If multiple prediction blocks are generated and combined, then prediction accuracy is improved, but processing operations increase
Solution Approach 1:
The patent segments the prediction process into distinct base hypothesis and additional hypothesis components. Each hypothesis generates its own prediction block independently using separate motion information from neighboring blocks. This segmentation allows parallel processing of multiple hypotheses without excessive interdependence, enabling the decoder to generate and combine prediction blocks efficiently while maintaining improved prediction accuracy through the combination of segmented hypothesis results.
3Device complexity
If motion information is coded using merge mode for additional prediction hypothesis, then signaling overhead is reduced, but motion information flexibility is constrained
Solution Approach 1:
The patent changes the parameters of motion information by adapting motion vectors and reference picture indices from neighboring blocks to form candidate motion information. This adaptation process transforms existing motion information parameters into suitable candidates for both base and additional hypotheses. By modifying parameters through adaptation rather than transmitting raw motion information, the patent reduces signaling overhead while maintaining flexibility through the adapted candidate lists that can be selected based on prediction performance.
Data Source
AI summary
An example device for decoding video data includes one or more processors configured to: generate a first prediction block for a current block of video data using a base inter-prediction mode; code a merge mode syntax element for a second prediction block representing an additional prediction hypothesis, the merge mode syntax element indicating whether motion information for the second prediction block is coded using merge mode; code the motion information for the second prediction block according to the merge mode syntax element, wherein to code the motion information, the one or more processors are configured to form a merge candidate list including merge candidates representing respective sets of uni-prediction motion information; generate the second prediction block for the current block of video data using the motion information; form a multi-hypothesis prediction block from the first and second prediction blocks; and decode the current block using the multi-hypothesis prediction block.