Sample-Wise Extrapolated Intra Prediction for Efficient Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in achieving high compression efficiency and maintaining video quality, particularly in methods like HEVC/H.265 and VVC/H.266, as they do not adequately utilize intra prediction techniques to reduce redundancy in video data.
Innovation Solution
Implementing sample-wise extrapolated intra prediction using n-tap extrapolation filters, which are selected and adapted through various models and classifiers to improve coding accuracy by using different shapes and prediction orders, and leveraging partially reconstructed blocks for more accurate predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional intra prediction methods (HEVC/VVC) are used, then device complexity is reduced, but coding accuracy and compression efficiency deteriorate
Solution Approach 1:
The patent applies parameter changes by using n-tap extrapolation filters with different tap counts (e.g., 5-tap, 7-tap, 9-tap filters) to modify the prediction process. The filter tap parameter is adjusted based on block characteristics, allowing the system to achieve higher coding accuracy by selecting appropriate filter complexity for different prediction scenarios without uniformly increasing device complexity across all blocks.
Solution Approach 2:
The patent implements dynamics by adaptively selecting different extrapolation filter types and tap counts based on block characteristics such as gradient calculations and variance measurements. This dynamic adaptation allows the prediction method to adjust its complexity level according to the specific block being processed, improving overall coding accuracy while maintaining reasonable device complexity through selective application.
2Measurement precision
If n-tap extrapolation filters are applied to all blocks, then coding accuracy improves, but computational cost and processing time increase
Solution Approach 1:
The patent applies local quality by selectively applying n-tap extrapolation filters only to specific blocks that benefit from them, rather than uniformly applying filters to all blocks. Blocks are classified based on their characteristics (e.g., gradient magnitude, variance), and filters are applied locally only where they provide meaningful accuracy improvement, thus maintaining processing speed while enhancing prediction accuracy for relevant blocks.
Solution Approach 2:
The patent implements partial action by applying extrapolation filters to only a subset of blocks that meet certain criteria, rather than applying them excessively to all blocks. This selective application based on block characteristics ensures that the computational cost is incurred only where necessary to improve accuracy, balancing processing speed and prediction quality.
3Productivity
If multiple filter shapes and prediction orders are used, then compression efficiency improves, but device complexity and implementation difficulty increase
Solution Approach 1:
The patent applies segmentation by dividing the filter selection process into distinct segments or categories based on block characteristics. Different filter shapes (e.g., horizontal, vertical, diagonal) and prediction orders are assigned to different block types or gradient directions, creating a segmented approach that manages complexity through structured classification rather than exhaustive search.
Solution Approach 2:
The patent implements preliminary action by pre-defining and pre-classifying different filter types and their appropriate application scenarios based on block characteristics. The classification criteria and filter mappings are established in advance, allowing the decoder to quickly select appropriate filters without complex real-time decision-making, thus improving compression efficiency while managing device complexity through pre-computed classification rules.
Data Source
AI summary
The various implementations described herein include methods and systems for coding video. In one aspect, a method includes receiving a video bitstream comprising a plurality of blocks that includes a current block; selecting a selected extrapolation filter from a set of extrapolation filters; deriving a prediction sample for the current block using an extrapolated intra prediction with the selected extrapolation filter; and reconstructing the current block using the derived prediction sample.