Generalized Bi-Prediction Syntax for HEVC B-Picture Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Modern video coding systems, particularly in B-pictures, often fail to provide optimum prediction, as they are limited to single or bi-directional prediction using one motion vector from each list, which can lead to increased compression noise and inefficiencies.
Innovation Solution
The implementation of a generalized bi-directional prediction (GBP) syntax allows both motion vectors to be selected from the same list, enabling improved prediction by averaging these vectors, thereby reducing compression noise and enhancing coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If bi-directional prediction uses one motion vector from each list, then prediction coverage is improved, but compression noise increases and prediction accuracy deteriorates
Solution Approach 1:
The invention changes the parameter of motion vector selection by allowing both motion vectors to be selected from the same reference list rather than requiring one from each list. This parameter change enables the averaging of two motion vectors from the same list, which reduces compression noise while maintaining prediction coverage, thereby resolving the contradiction between prediction accuracy and compression noise.
2Device complexity
If single prediction is used in B-frames, then encoding complexity is reduced, but prediction efficiency deteriorates
Solution Approach 1:
The invention introduces dynamic flexibility to the prediction process by allowing the encoder to choose between single prediction and bi-directional prediction with both vectors from the same list based on rate-distortion cost. This dynamic adaptation enables the system to achieve better prediction efficiency when needed while maintaining manageable encoding complexity by using simple single prediction when sufficient.
3Ease of operation
If bi-directional prediction with one MV from each list is fixed, then decoding simplicity is maintained, but prediction optimality deteriorates
Solution Approach 1:
The invention creates a universal prediction framework that encompasses both the traditional approach (one MV from each list) and the new approach (both MVs from the same list) within a single bi-directional prediction structure. This multi-functionality allows the system to maintain decoding simplicity while achieving prediction optimality by selecting the appropriate motion vector combination based on rate-distortion cost, thereby resolving the contradiction between decoding simplicity and prediction optimality.
Data Source
AI summary
An enhanced prediction direction method and encoding syntax for B-pictures in state-of-the-art video compression. In bi-directional prediction, instead of using one motion vector (MV) from list 0 and one from list 1, both may be alternatively drawn from list 0 or list 1, and this information communicated to a decoder similarly configured for recognizing the generalized syntax.