Intra Prediction Using Reduced Neighbor Samples and Affine Transforms
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in achieving efficient intra-prediction for video blocks, particularly in HEVC and its successors, due to high computational complexity and inefficiencies in using neural networks for non-linear transformations.
Innovation Solution
Implementing a decoder and encoder that utilize linear or affine transforms with neighboring sample reduction by grouping neighboring samples into groups and applying downsampling or averaging, followed by a linear or affine transformation to predict block values, with specific transformation sets based on block dimensions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If neural networks are used for non-linear transformations in intra-prediction, then prediction accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent replaces complex neural network transformations with simpler, disposable-like linear and affine transformations that achieve sufficient prediction accuracy without the high computational cost of neural networks. These simpler transforms are applied once and discarded, avoiding the need for complex, reusable neural network models.
Solution Approach 2:
The patent changes the transformation parameters from complex neural network weights to simpler linear and affine transformation matrices. This parameter simplification reduces computational complexity while maintaining adequate prediction accuracy for intra-prediction in video coding.
2Measurement precision
If all neighbouring samples are used for prediction, then prediction accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent extracts and processes only the most relevant neighbouring samples by applying linear and affine transformations that selectively weigh and combine boundary samples. This extraction approach maintains prediction accuracy by focusing on the most informative samples rather than processing all neighbouring samples equally.
Solution Approach 2:
The patent applies partial action by using linear and affine transformations that process a subset of neighbouring samples with appropriate weighting, rather than fully processing all neighbouring samples. This partial processing reduces computational complexity while maintaining sufficient prediction accuracy.
3Measurement precision
If complex transformation matrices are applied to all block sizes, then prediction accuracy is improved, but device complexity increases
Solution Approach 1:
The patent applies local quality by selecting different linear and affine transformation matrices tailored to specific block sizes. Each block size receives an optimized transformation matrix that is locally adapted to its dimensions, avoiding the need for a single complex universal transformation matrix while maintaining prediction accuracy across different block sizes.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
There are provided techniques for encoding/decoding of video signals, e.g. implemented in decoders, encoder, methods and non-transitory storage units storing instructions to perform the methods. A decoder or encoder may be configured to predict a predetermined block of the picture using a plurality of neighbouring samples by reducing the plurality of neighbouring samples to obtain a reduced set of samples values lower, in number of samples, than compared to the plurality of neighbouring samples, subjecting the reduced set of sample values to a linear or affine linear transformation to obtain predicted values for predetermined samples of the predetermined block.