Intra Prediction Interpolation Filters for Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies face challenges in reducing data transmission efficiency due to limitations in interpolation filtering, particularly with longer-tap filters and reference sample management, which lead to increased complexity and prediction errors.
Innovation Solution
The use of dynamic reference sample buffers and multiple interpolation filters based on block characteristics and prediction modes to select the optimal filter taps and sample selection, reducing the need for clipping operations and rounding errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If longer-tap interpolation filters are used to improve prediction accuracy, then manufacturing precision is improved, but device complexity increases
Solution Approach 1:
The patent implements dynamic selection of interpolation filter tap lengths based on block characteristics and prediction modes. The system adapts between 2-tap, 4-tap, and 6-tap filters depending on the specific coding context, allowing longer filters to be used only when beneficial for accuracy while avoiding unnecessary complexity in other cases.
Solution Approach 2:
The patent changes the parameter of filter tap length dynamically based on block size, shape, and prediction mode. By adjusting this key parameter adaptively rather than using a fixed long filter, the system achieves high prediction accuracy where needed while maintaining lower complexity in suitable cases.
2Manufacturing precision
If more reference samples are stored in the reference buffer to support longer-tap filters, then manufacturing precision is improved, but device complexity increases
Solution Approach 1:
The reference buffer size and content are dynamically adjusted based on the selected filter tap length and block characteristics. When a 6-tap filter is chosen, the buffer is populated with sufficient reference samples; when shorter filters are used, the buffer management is simplified accordingly.
Solution Approach 2:
The reference buffer is managed in segments corresponding to different filter requirements. The system divides reference sample storage and retrieval operations into segments based on the active filter configuration, making the complex buffer management more tractable and efficient.
3Reliability
If clipping operations are performed to handle boundary reference samples, then reliability is improved, but manufacturing precision deteriorates due to rounding errors
Solution Approach 1:
The patent performs preliminary extension of reference samples beyond the block boundaries before clipping is needed. By pre-computing and storing extended reference samples in the buffer, the system avoids the need for clipping operations during prediction, thereby eliminating rounding errors while maintaining reliability.
Solution Approach 2:
Extended reference samples act as an intermediary between the actual block boundaries and the filter requirements. These intermediate samples allow the filter to operate with full precision without directly clipping boundary samples, thus preserving prediction accuracy.
4Adaptability or versatility
If multiple interpolation filters are implemented to adapt to different block characteristics, then adaptability is improved, but device complexity increases
Solution Approach 1:
The system dynamically selects from multiple filter types (2-tap, 4-tap, 6-tap) based on real-time analysis of block characteristics including size, shape, and prediction mode. This dynamic adaptation provides versatility while avoiding the need to permanently implement all filter variants with full complexity.
Solution Approach 2:
The system changes the filter parameter (tap length) based on block characteristics rather than implementing fundamentally different filter structures. This parameter-based adaptation achieves versatility with relatively simple filter implementations that can be configured dynamically.
Data Source
AI summary
Techniques are described in which a video coder is configured to determine, using one or more characteristics of an interpolation filter, a number of reference samples to be stored at a reference buffer. The video coder is further configured to generate a plurality of values corresponding to the number of reference samples in the reference buffer. The video coder is further configured to generate prediction information for intra-prediction using the interpolation filter and the plurality of values. The video coder is further configured to reconstruct the block of video data based on the prediction information.