Video Encoding Temporal Prediction Overwriting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video compression standards, such as AVC, have limitations in prediction methods that restrict scalability and compatibility, particularly in achieving higher frame rates without significant bit-rate increases or complex changes to existing procedures.
Innovation Solution
The method involves bypassing the conventional temporal prediction dogma by overwriting older, less important predictions in the enhancement encoder with separate predictions from the base stream, allowing for more flexible and efficient use of memory space without requiring major changes to the standard, enabling scalable compression and higher frame rates like 75Hz.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple temporal predictions are stored in memory for video compression, then prediction accuracy is improved, but memory usage and device complexity increase
Solution Approach 1:
The patent applies the principle of discarding and recovering by selectively overwriting old prediction frames in the reference buffer with new prediction frames from the enhancement layer. Instead of maintaining all historical prediction frames indefinitely, the system discards less important old predictions and recovers memory space for more valuable new predictions, achieving a balance between prediction accuracy and memory usage
Solution Approach 2:
The patent implements dynamic management of the reference buffer by adjusting which prediction frames are retained and which are overwritten based on their importance and age. The system dynamically adapts the prediction strategy by selecting between different prediction sources (temporal predictions vs. enhancement layer predictions) depending on current buffer state and prediction quality requirements
2Measurement precision
If more frames are stored in buffer for temporal prediction, then prediction quality improves, but bit rate increases
Solution Approach 1:
The patent segments the prediction process into two independent parts: base layer prediction and enhancement layer prediction. By separating these functions, the system can use the reference buffer efficiently for base layer temporal predictions while using enhancement layer predictions for additional quality improvement, avoiding the need to store excessive frames solely for quality enhancement
Solution Approach 2:
The patent introduces enhancement layer prediction frames as an intermediary mechanism to improve prediction quality without requiring additional temporal frames in the buffer. These enhancement predictions act as mediators that provide additional prediction information without increasing the temporal buffer size or base layer bit rate
3Adaptability or versatility
If conventional temporal prediction is used in enhancement encoder, then compatibility with existing standards is maintained, but coding efficiency is limited
Solution Approach 1:
The patent merges conventional temporal prediction with enhancement layer prediction by combining results from both prediction sources. The enhancement encoder maintains compatibility with existing standards through temporal prediction while simultaneously incorporating enhancement layer predictions to improve coding efficiency, achieving both compatibility and enhanced performance
Solution Approach 2:
The patent creates a composite prediction approach by combining multiple prediction sources (temporal predictions from reference buffer and enhancement layer predictions) into a unified prediction mechanism. This composite approach maintains the structural compatibility of conventional temporal prediction while integrating additional prediction information to improve overall coding efficiency
Data Source
Figure 1a
Figure 1b
Figure 1c
AI summary
Some video compression standards use multiple temporal predictions. One or more of the oldest temporal predict ions are overwritten with another prediction. A prediction used in an enhancement encoder is in embodiment overwritten by a prediction produced in a base stream encoder.