Intersecting Motion Vector Prediction for Temporal Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current motion vector prediction techniques in video coding are inefficient for all situations, particularly when deriving an MVP candidate from a constructive motion vector or zero motion vector, which fails to effectively model temporal information for inter predicted blocks.
Innovation Solution
The techniques utilize intersecting motion vectors, temporal motion vector prediction, and reference picture derivation, incorporating both spatial and temporal neighbors, and motion vectors from already decoded pictures to improve MVP derivation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If motion vector prediction is derived from constructive motion vector or zero motion vector, then the prediction process is simple, but the temporal information modeling efficiency deteriorates
Solution Approach 1:
The patent merges spatial neighbor motion vectors with temporal reference picture information to create intersecting motion vectors. This combination allows the system to leverage both spatial and temporal dimensions for prediction, resolving the contradiction by integrating multiple data sources rather than relying on a single simple method.
Solution Approach 2:
The invention transitions from one-dimensional prediction (using only constructive or zero motion vectors) to two-dimensional prediction by incorporating both spatial neighbor information and temporal reference pictures. This dimensional expansion enables better temporal information modeling while maintaining computational feasibility.
2Ease of manufacture
If current MVP derivation techniques are used, then the encoding process is straightforward, but bandwidth efficiency and decoding quality deteriorate
Solution Approach 1:
The patent introduces intersecting motion vectors as an intermediary that bridges spatial and temporal prediction methods. This intermediary mechanism enables the system to achieve both straightforward encoding and high decoding quality by providing a unified approach that leverages multiple prediction sources.
Solution Approach 2:
The invention creates a composite prediction system that combines multiple motion vector derivation methods (spatial neighbors, temporal references, constructive vectors) into a unified MVP derivation process. This composite approach maintains encoding simplicity while significantly improving bandwidth efficiency and decoding quality.
Data Source
AI summary
Example methods, devices, and computer-readable media are described. An example device includes one or more processors configured to determine, for each of a plurality of subblocks of a current block of a current picture of video data, a corresponding intersecting motion vector list. Each corresponding intersecting motion vector list includes at least one intersecting motion vector. The at least one intersecting motion vector includes a source motion vector associated with a first reference picture and a destination motion vector associated with a second reference picture. A source block is located in the first reference picture and has a predictor that is a subset of a destination block located in the second reference picture. The one or more processors are configured to determine an intersecting motion vector for the current block based on the corresponding intersecting motion vector lists and decode the current block based on the intersecting motion vector.