Sub-Block Intra Prediction Using Filtered Reference Pixels
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image encoding and decoding technologies require improvements in performance and efficiency, particularly in intra prediction methods, reference pixel construction, filtering, and interpolation.
Innovation Solution
An image encoding/decoding method and apparatus that constructs an intra prediction mode candidate group, determines a reference pixel, performs filtering or interpolation based on specific criteria, and conducts intra prediction in sub-block units.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing image encoding/decoding methods are used, then basic encoding functionality is provided, but encoding performance and efficiency are insufficient
Solution Approach 1:
The patent divides the image block into multiple sub-blocks for independent intra-prediction processing. This segmentation allows more flexible prediction mode selection and reference pixel construction for each sub-block, improving encoding efficiency and performance by adapting to local image characteristics rather than applying a uniform approach to the entire block.
Solution Approach 2:
The patent applies different prediction modes and reference pixel construction methods to different sub-blocks based on their specific characteristics. By allowing variable prediction modes (e.g., planar, angular, diagonal) and selective reference pixel filtering/interpolation in different regions, the system optimizes encoding performance for each local area rather than using a one-size-fits-all approach.
2Measurement precision
If reference pixel filtering is applied, then prediction accuracy is improved, but processing time increases
Solution Approach 1:
The patent applies reference pixel filtering selectively rather than universally. Filtering is applied only to reference pixels that are actually needed for the chosen prediction mode, and the degree of filtering (e.g., mean filtering vs. no filtering) is adjusted based on the specific prediction mode and block characteristics. This partial application reduces unnecessary processing time while maintaining accuracy where needed.
Solution Approach 2:
The patent dynamically adjusts the reference pixel filtering process based on the selected prediction mode. For example, certain prediction modes may require filtering while others do not, and the system adapts its behavior accordingly. This dynamic approach allows the system to optimize between accuracy and speed based on the specific encoding conditions.
3Measurement precision
If interpolation is performed on reference pixels, then prediction precision is improved, but computational complexity increases
Solution Approach 1:
The patent applies interpolation selectively to reference pixels based on the prediction mode and sub-block characteristics. Interpolation is used only when necessary to achieve accurate prediction, such as when fractional pixel sampling is required, rather than applying it uniformly to all reference pixels. This reduces overall computational complexity while maintaining precision where needed.
4Adaptability or versatility
If intra prediction is performed in sub-block units, then encoding flexibility is improved, but processing steps increase
Solution Approach 1:
The patent segments the prediction process into sub-block units, allowing independent prediction mode selection for each segment. This increases flexibility by enabling adaptive prediction strategies tailored to local content, while the segmentation itself provides a structured framework that manages complexity through systematic processing of smaller units.
Data Source
AI summary
A method and apparatus for encoding/decoding images according to the present invention may configure a prediction mode candidate group on a screen of a target block, determine a reference pixel for a prediction mode on the screen of the target block, perform at least one of filtering and interpolation on the reference pixel, and perform prediction on the screen of the target block on the basis of the prediction mode candidate group on the screen and the reference pixel.