Video Decoder Composed Prediction for Lower Signaling Overhead
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 by enhancing block-based predictive video coding techniques.
Innovation Solution
Implement a composed prediction signal by combining multiple primitive prediction modes, allowing more than two prediction signals per block, with controlled composition weights and merge candidate restrictions to reduce noise and signaling overhead.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple primitive prediction modes are combined to form composed prediction signals, then prediction accuracy is improved, but signaling overhead increases
Solution Approach 1:
The prediction process is segmented into multiple primitive prediction modes (e.g., planar, DC, gradient predictions) that can be independently generated and then combined. Each primitive prediction mode processes different aspects of the block independently, allowing selective combination to achieve high accuracy while managing signaling overhead through modular processing.
Solution Approach 2:
Multiple prediction signals from different primitive prediction modes are merged into a single composed prediction signal using weighted combination. This merging process integrates the strengths of different prediction modes (e.g., planar for smooth regions, gradient for edges) to improve overall prediction accuracy while the weights control the contribution of each mode.
2Reliability
If more than two prediction signals are combined per block, then noise reduction is improved, but device complexity increases
Solution Approach 1:
The system allows combining more than two prediction signals (excessive action) to enhance noise reduction, but implements selective combination where not all possible prediction modes are always activated. The decoder can choose to combine only the necessary number of prediction signals based on block characteristics, reducing complexity while maintaining noise reduction benefits where needed.
Solution Approach 2:
The number of prediction signals to be combined is made variable rather than fixed. The system can dynamically adjust the number of combined prediction signals based on block complexity, motion characteristics, and other parameters, allowing noise reduction to be enhanced when beneficial while limiting complexity increase when unnecessary.
3Measurement precision
If composed prediction with flexible composition weights is used, then prediction quality is improved, but ease of operation deteriorates
Solution Approach 1:
The system implements self-service through automatic weight determination mechanisms where composition weights are either derived from block characteristics (e.g., variance, gradient) or determined by simple rules rather than requiring complex manual configuration. This allows high prediction quality through flexible weighting while maintaining ease of operation by eliminating the need for manual parameter tuning.
Solution Approach 2:
The composition weights are computed on-the-fly using simple algorithms rather than being stored as complex pre-configured parameters. This disposable computation approach provides flexible weighting for high prediction quality while keeping the system easy to operate by avoiding the need for manual weight configuration or complex stored parameter sets.
Data Source
AI summary
A video decoder for decoding a video from a data stream using block-based predictive decoding is arranged to derive, using inter prediction, a first prediction, p1, for a current block, derive, using intra prediction, a second prediction, p2, for the current block determine a scalar weighting parameter, α, based at least in part on a look up table using prediction parameter information of blocks neighboring the current block, determine a prediction signal, q, for the current block as a weighted sum, of the first prediction, p1, and the second prediction, p2, using, at least in part, the scalar weighting parameter, α, and decode the current block using the prediction signal, q.