Intra Reference Filter Bilateral Smoothing for Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face inefficiencies in intra prediction, particularly in regions with rich slanting texture patterns and for non-square blocks, where current smoothing filters can reduce compression efficiency and performance.
Innovation Solution
The use of bilateral filters with mode-dependent intra smoothing (MDIS) techniques to determine filter coefficients based on distance and sample value differences between reference samples, improving intra prediction by applying these filters to produce filtered reference samples for enhanced compression performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Stability of the object's composition
If conventional smoothing filters are applied to reference samples for intra prediction, then prediction smoothness is improved, but compression efficiency deteriorates in regions with sharp edges and non-square blocks
Solution Approach 1:
The patent applies different filtering operations to different regions of the reference samples based on their spatial location and relationship to the current block. Specifically, it applies filtering to reference samples that are not adjacent to the current block while excluding filtering for reference samples that are adjacent to the current block. This local differentiation allows the system to maintain prediction smoothness in homogeneous regions while preserving sharp edges and geometric accuracy in regions adjacent to the current block, thereby resolving the contradiction between smoothness and compression efficiency.
2Device complexity
If filter coefficients are applied to all reference samples uniformly, then processing complexity is reduced, but prediction accuracy deteriorates for non-square blocks and regions with slanting textures
Solution Approach 1:
The patent differentiates the treatment of reference samples based on their spatial relationship to the current block. It specifically excludes filtering operations for reference samples that are adjacent to the current block, while applying filtering to non-adjacent reference samples. This local differentiation enables the system to maintain simple processing for most cases while achieving high prediction accuracy for non-square blocks and regions with slanting textures by selectively applying filters only where needed, thus resolving the contradiction between processing complexity and prediction accuracy.
3Speed
If intra prediction modes are applied without considering block geometry, then coding speed is maintained, but coding efficiency deteriorates for non-square blocks
Solution Approach 1:
The patent introduces geometry-aware filtering by differentiating the treatment of reference samples based on their spatial relationship to the current block. The system applies filtering operations selectively to non-adjacent reference samples while excluding adjacent reference samples from filtering. This geometry-aware approach allows the system to maintain fast coding speeds through efficient filtering operations while significantly improving coding efficiency for non-square blocks by preserving geometric accuracy in the prediction process.
Data Source
AI summary
A method of decoding video data that includes receiving a current block of video data encoded using an intra prediction mode, determining residual video data for the current block of video data, determining reference samples of the current block of video data, determining filter coefficients for a bilateral filter based on a distance between the reference samples and neighboring reference samples and based on a sample value difference between the reference samples and the neighboring reference samples, applying the bilateral filter with the determined filter coefficients to the determined reference samples to produce filtered reference samples, generating a prediction block using the filtered reference samples, and adding samples of the prediction block to the determined residual video data to produce a decoded block of video data.