Composed Video Prediction With Restricted Merge for Coding Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video codecs, such as AVC/H.264 and HEVC/H.265, while effective, can further improve coding efficiency in block-based predictive video coding by enhancing the prediction modes to reduce prediction residuals and side information.
Innovation Solution
Implement a video decoder and encoder that support composed prediction signals by combining multiple primitive prediction modes, allowing more than two prediction signals per block, with controlled composition weights and merge candidate restrictions to optimize coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If multiple primitive prediction modes are combined to form composed prediction signals, then prediction accuracy and coding efficiency are improved, but device complexity increases
Solution Approach 1:
The prediction process is segmented into multiple independent primitive prediction modes (first prediction signal, second prediction signal, etc.), each generated using different prediction methods (intra prediction, inter prediction, etc.). These segmented prediction signals are then combined through composition to form the final composed prediction signal, allowing the system to achieve high prediction accuracy while maintaining manageable complexity through modular design
Solution Approach 2:
Multiple primitive prediction signals generated from different prediction modes are merged/combined to form a composed prediction signal. The combining process uses composition weights to weight each primitive prediction signal, and the weighted signals are summed to produce the final prediction. This merging approach leverages the strengths of different prediction methods to improve overall prediction accuracy while the patent manages complexity through efficient combination mechanisms
2Productivity
If composed prediction signals with multiple primitive predictions are used, then coding efficiency is improved, but signaling overhead increases
Solution Approach 1:
The system uses a variable number of primitive prediction signals (K) where K can be 1, 2, or more, depending on what is needed for optimal prediction. This partial action approach allows the encoder to use only as many prediction modes as necessary for a given block, avoiding the excessive signaling overhead that would result from always using the maximum number of prediction modes, while still achieving improved coding efficiency when multiple modes are beneficial
Solution Approach 2:
The patent employs parameter changes in the form of composition weights that can be signaled at different granularities (picture level, sequence level, or block level). By allowing these weights to be signaled at coarser granularities, the system reduces the amount of signaling overhead required while still enabling precise control over the composition of multiple prediction signals, thus improving coding efficiency without proportionally increasing signaling overhead
3Loss of energy
If prediction modes are enhanced to reduce prediction residuals, then coding efficiency is improved, but computational complexity increases
Solution Approach 1:
The prediction process is divided into multiple independent primitive prediction operations, each handling a specific aspect of prediction (e.g., spatial prediction, temporal prediction). This segmentation allows the system to compute predictions in parallel or in a structured manner, reducing the overall computational complexity compared to a single complex prediction operation, while still achieving lower prediction residuals through the combination of multiple specialized predictions
Solution Approach 2:
The system performs preliminary actions by generating multiple primitive prediction signals before final composition. Each primitive prediction can be pre-computed using efficient algorithms, and then the results are combined. This preliminary action approach allows for optimized computation of individual prediction components, reducing the overall computational burden while achieving enhanced prediction accuracy through the combination of pre-computed predictions
Data Source
AI summary
A method of decoding a video from a data stream using block-based predictive decoding using a video decoder, includes, for a predetermined block, reading first prediction information from the data stream, determining, based on the first prediction information, a first prediction signal (p1), deriving a number K from the data stream, determining K further prediction signals (p2 . . . pK+1) and for each of the K further prediction signals, a composition weight, and predicting the predetermined block based on the first prediction signal and the K further prediction signals and the composition weights therefor.