Adaptive Intra Interpolation Filtering for Angular Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in efficiently utilizing intra interpolation filters for angular prediction modes, leading to suboptimal compression and reconstruction quality.
Innovation Solution
A method for selecting and applying intra interpolation filters based on neighboring reconstructed samples, including types and number of taps, to enhance angular intra prediction modes, using techniques such as neural networks for filter selection and minimizing prediction errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a fixed intra interpolation filter is used for angular prediction modes, then the device complexity is reduced, but the reconstruction quality and compression efficiency deteriorate
Solution Approach 1:
The patent implements dynamic filter selection by evaluating multiple intra interpolation filters (different types and tap configurations) based on neighboring reconstructed samples and selecting the optimal filter for each angular prediction mode. This transforms the static filter application into a dynamic adaptive process that improves reconstruction quality while maintaining manageable complexity through systematic evaluation criteria.
Solution Approach 2:
The patent changes the parameters of the intra interpolation filter by considering different filter types (e.g., linear, cubic, spline) and different numbers of taps (e.g., 2-tap, 4-tap, 6-tap, 8-tap) based on the characteristics of neighboring reconstructed samples. This parameter variation allows the system to adapt to different image content characteristics and improve compression efficiency.
2Manufacturing precision
If multiple intra interpolation filters are evaluated and selected based on prediction error, then the reconstruction quality improves, but the computational complexity increases
Solution Approach 1:
The patent applies partial evaluation by considering a predefined set of intra interpolation filters with different types and tap configurations rather than evaluating all possible filters. This selective evaluation approach achieves sufficient prediction accuracy improvement while controlling the computational complexity through a managed set of candidate filters.
Solution Approach 2:
The patent implements feedback mechanisms by calculating prediction errors for each candidate filter based on neighboring reconstructed samples and using these errors to select the optimal filter. The prediction error serves as feedback that guides the filter selection process, ensuring that the chosen filter minimizes the discrepancy between predicted and actual samples.
3Adaptability or versatility
If neural network-based filter selection is used, then the adaptability to different image content improves, but the processing time and computational resources increase
Solution Approach 1:
The patent performs preliminary analysis by evaluating multiple intra interpolation filters using neighboring reconstructed samples before final prediction. This preliminary evaluation establishes a foundation for selecting the most appropriate filter, enabling the system to adapt to different image content characteristics while providing a systematic approach to filter selection that balances adaptability with processing efficiency.
Data Source
AI summary
An apparatus includes processing circuitry that receives, from a bitstream including a current block in a picture, coding information of the bitstream. The coding information indicates that the current block is coded in an angular intra prediction mode with an intra interpolation filter. The processing circuitry applies each of a predefined set of intra interpolation filters to neighboring reconstructed samples within N adjacent lines from a boundary of the current block. The processing circuitry selects one intra interpolation filter from the predefined set of intra interpolation filters based a prediction error associated with the each of the predefined set of intra interpolation filters and predicts a sample in the current block using the angular intra prediction mode using the selected one intra interpolation filter. The processing circuitry reconstructs the current block based on the predicted sample.