Planar Prediction Mode Reference Sample Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards, such as H.265/HEVC, face inefficiencies in the planar prediction mode due to fake texture characteristics introduced by padding neighboring samples, leading to reduced prediction accuracy and efficiency, especially for larger blocks where spatial correlation among samples decreases.
Innovation Solution
The proposed method improves video encoding and decoding by selecting a first set of reconstructed neighboring samples and determining a prediction value through interpolation, using a weighted sum of samples aligned horizontally and vertically with the prediction sample, and positioned on the opposite side, thereby enhancing prediction accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If padding neighboring samples is used in planar prediction mode, then the prediction process can be simplified, but fake texture characteristics are introduced reducing prediction accuracy
Solution Approach 1:
The patent extracts and removes the problematic padded samples from the reference sample set. Specifically, it identifies samples that are padded (fake texture sources) and excludes them from being used in the planar prediction calculation, thereby eliminating the source of prediction errors while maintaining the simplified prediction process for remaining valid samples.
Solution Approach 2:
The patent applies different quality criteria to different reference samples. Instead of treating all neighboring samples uniformly, it selectively validates each sample to determine whether it is a genuine reconstructed sample or a padded sample, applying local quality assessment to ensure only high-quality samples contribute to the prediction.
2Measurement precision
If more reference samples are used for interpolation, then prediction accuracy improves, but computational complexity increases
Solution Approach 1:
The patent performs preliminary validation of reference samples before they are used in interpolation. By pre-identifying and marking valid versus invalid samples, and pre-calculating their positions and weights, the system avoids complex runtime decisions during the actual prediction process, thus maintaining high accuracy with controlled complexity.
Solution Approach 2:
The patent dynamically adjusts the set of reference samples based on block size and position parameters. For different block configurations, it changes which samples are included in the interpolation, optimizing the balance between accuracy and complexity for each specific case rather than using a fixed complex algorithm for all cases.
3Productivity
If traditional planar prediction with padding is used, then encoding speed is maintained, but prediction performance degrades for larger blocks
Solution Approach 1:
The patent introduces dynamic adaptation in the planar prediction process. The set of reference samples is dynamically selected based on the current block's size, position, and available neighboring samples. This dynamic approach allows the algorithm to optimize for larger blocks where spatial correlation is weaker, improving performance without sacrificing encoding speed through excessive complexity.
Data Source
AI summary
Devices, systems and methods for deriving intra prediction samples when the intra prediction mode for a coding block is the planar mode are described. In an aspect, a method for video coding includes selecting a first set of reference samples that are reconstructed neighboring samples of a current block, and determining a prediction value for a prediction sample of the current block by interpolating at least one of the first set and a second set of reference samples, where a reference sample of the second set of reference samples is based on a weighted sum of a first sample and a second sample from the first set of reference samples, and where the reference sample is aligned horizontally with the first sample, aligned vertically with the second sample, and positioned on an opposite side of the prediction sample with respect to either the first or the second sample.