Directional Intra Prediction Using Fitting Plane and Secondary Reference Samples
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
State-of-the-art video coding standards, such as ITU-T H.265/HEVC, face challenges in effectively handling gradual changes in intensity within video blocks, particularly in textured or edge regions, as directional intra-prediction modes fail to compensate for both directional and gradient components.
Innovation Solution
A novel method for generating secondary reference samples using a combination of gradient and directional prediction signals, where gradient signals are modeled via Multiple Linear Regression (MLR) and directional signals are projected onto secondary reference samples based on the intra-prediction mode, allowing for improved prediction across different components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If directional intra-prediction modes are used, then prediction accuracy for directional structures is improved, but the ability to handle gradual intensity changes in textured or edge regions deteriorates
Solution Approach 1:
The patent combines gradient prediction and directional prediction into a unified hybrid prediction mode. The gradient component captures gradual intensity changes while the directional component handles edge structures, allowing the prediction system to simultaneously address both requirements that previously conflicted with each other.
Solution Approach 2:
The prediction signal is constructed as a composite of two distinct components: a gradient signal that models gradual intensity variations and a directional signal that captures edge structures. This composite approach allows the system to leverage the strengths of both prediction types while mitigating their individual weaknesses.
2Productivity
If conventional intra-prediction methods are used, then computational complexity is reduced, but coding efficiency deteriorates due to inability to compensate for both gradient and directional components
Solution Approach 1:
The prediction process is segmented into distinct stages: gradient calculation from reference samples, directional signal generation based on intra-prediction mode, and combination of these signals. This segmentation allows each component to be optimized independently while maintaining overall efficiency.
Solution Approach 2:
The gradient signal is calculated in advance from available reference samples before the actual prediction process. This preliminary calculation of the gradient component allows the main prediction stage to focus only on combining pre-computed gradient information with directional information, reducing overall computational burden.
Data Source
AI summary
A video coding block is intra-predicted on the basis of a plurality of primary reference samples from a first set of neighboring video coding blocks. A fitting plane is determined on the basis of the plurality of primary reference samples and defines a plurality of fitting samples. For a selected directional intra prediction mode, a plurality of secondary reference samples are generated on the basis of the plurality of primary reference samples and the plurality of fitting samples. The secondary reference samples are located in a second set of neighboring video coding blocks. The sample values of the current video coding block are predicted on the basis of the plurality of primary reference sample values and the plurality of secondary reference sample values.