Temporal Motion Vector Candidates for Memory-Efficient Inter Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image encoding and decoding technologies face inefficiencies in handling high-resolution and high-definition images, particularly in inter prediction methods that require storing reference pictures in memory and suffer from error resilience issues.
Innovation Solution
An inter prediction method that derives reference motion information from reference pictures by counting occurrence frequencies, performing median operations, sub-sampling, and grouping motion information to improve encoding/decoding efficiency and error resilience.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If reference pictures are stored in memory for inter prediction, then motion compensation can be performed, but memory requirements increase
Solution Approach 1:
The patent extracts only the necessary motion information (motion vectors, reference picture indices) from reference pictures rather than storing entire reference pictures in memory. This extraction approach maintains inter prediction accuracy while significantly reducing memory requirements by storing only essential motion parameters.
Solution Approach 2:
The patent segments motion information into discrete components (motion vectors, reference picture indices, prediction directions) that can be stored and transmitted separately. This segmentation allows efficient memory usage by storing only relevant motion parameters rather than complete picture data.
2Productivity
If motion information from reference pictures is used for inter prediction, then encoding efficiency improves, but error propagation risk increases
Solution Approach 1:
The patent creates a simplified copy of motion information from reference pictures, storing only essential parameters (motion vectors, reference indices) rather than complete picture data. This copying approach maintains encoding efficiency while reducing error propagation by limiting the amount of data that can propagate errors.
Solution Approach 2:
The patent changes the parameters being stored from complete picture data to simplified motion parameters (motion vectors, reference picture indices). This parameter transformation maintains prediction accuracy while improving error resilience by reducing the complexity and potential error sources in the prediction process.
3Measurement precision
If detailed motion information is transmitted, then prediction accuracy improves, but bit transmission increases
Solution Approach 1:
The patent extracts only the most essential motion parameters (motion vectors, reference picture indices, prediction directions) needed for accurate prediction, excluding redundant information. This extraction maintains prediction accuracy while minimizing bit transmission requirements.
Solution Approach 2:
Instead of transmitting complete picture data and deriving motion information at the decoder, the patent inverts the approach by extracting and transmitting only motion parameters, which are then used to reconstruct prediction information at the decoder side.
Data Source
AI summary
An inter prediction method according to the present invention comprises: a step for deriving reference motion information related to a unit to be decoded in a current picture; and a step for performing motion compensation for the unit to be decoded, using the reference motion information that has been derived. According to the present invention, image encoding/decoding efficiency can be enhanced.