Intra Prediction Interpolation Filter for Video Coding Memory
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding standards face challenges in balancing bandwidth requirements and video quality, especially with the increasing complexity of intra prediction modes and the need for more efficient memory management in directional intra prediction.
Innovation Solution
The method involves a mapping process to simplify intra prediction calculations, selecting a subpixel interpolation filter based on the subpixel offset, and determining the size of the main reference side according to the intra-prediction mode and interpolation filter length, thereby optimizing memory efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If more reference samples are stored for directional intra prediction, then prediction accuracy is improved, but memory requirements increase
Solution Approach 1:
The patent dynamically adjusts the number of reference samples stored based on the intra-prediction mode parameter. For non-directional modes, only a minimal set of reference samples is stored, while for directional modes, additional reference samples are stored. This parameter-based adaptation resolves the contradiction by matching memory usage to actual prediction needs.
Solution Approach 2:
The reference sample buffer size is made dynamic rather than static. The system adapts the buffer capacity according to the specific prediction mode being used, allocating more memory only when directional prediction is required. This dynamic approach allows the system to maintain high prediction accuracy when needed while minimizing memory usage during non-directional prediction.
2Measurement precision
If complex intra prediction modes are implemented, then video quality is improved, but device complexity increases
Solution Approach 1:
The patent segments the intra-prediction process into distinct modes (directional and non-directional), each with its own optimized reference sample handling. This segmentation allows the system to apply complex processing only when necessary (directional modes) while using simpler processing for other cases, thereby maintaining video quality without uniformly increasing device complexity.
Solution Approach 2:
Different processing approaches are applied to different prediction modes. Directional intra-prediction modes receive enhanced treatment with additional reference samples and more complex calculations, while non-directional modes use simpler processing. This local quality approach ensures high video quality where needed without unnecessarily complicating the overall system.
3Reliability
If reference sample buffer size is increased, then all necessary samples are available for prediction, but memory efficiency decreases
Solution Approach 1:
The reference sample buffer size is changed as a parameter based on the prediction mode. The system calculates the appropriate buffer size dynamically, increasing it only when directional prediction modes require additional samples. This parameter adaptation ensures sample availability when needed while maintaining memory efficiency during non-directional prediction.
Solution Approach 2:
The buffer size transitions from a static fixed value to a dynamic value that adapts to prediction requirements. The system monitors the prediction mode and adjusts the buffer capacity accordingly, ensuring that all necessary samples are available for directional prediction while avoiding unnecessary memory allocation for non-directional prediction, thus resolving the contradiction between reliability and memory efficiency.
Data Source
AI summary
Provided are a method, an apparatus, a computer program product and a non-transitory computer-readable medium for video coding. The method comprises performing an intra-prediction process of a block comprising samples to be predicted, wherein an interpolation filter is applied to reference samples of the block during the intra-prediction process of the block. The interpolation filter is selected on the basis of a subpixel offset between the reference samples and the samples to be predicted, and a size of a main reference side used in the intra-prediction process is determined according to a length of the interpolation filter and an intra-prediction mode that provides, out of a set of available intra-prediction modes, the greatest non-integer value of said subpixel offset, wherein the main reference side comprises the reference samples.