Intra Prediction Filtering for Accurate Image Block Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image compression technologies face challenges in achieving high efficiency for high-resolution and high-quality images, particularly in accurately predicting pixel values within current pictures using intra-prediction methods.
Innovation Solution
An image decoding method that determines a reference block for intra prediction, generates prediction samples by applying filters to reference and neighboring samples, and reconstructs blocks based on these samples, using templates and filters adjusted by sampling rates and offsets to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional intra-prediction methods are used, then the encoding process is simple, but the prediction accuracy for high-resolution images is insufficient
Solution Approach 1:
The current block is divided into multiple sub-blocks, and different prediction modes are applied to different sub-blocks. This segmentation allows the patent to achieve high prediction accuracy for each sub-block while managing overall encoding complexity through selective application of complex filtering operations.
Solution Approach 2:
The patent applies complex filtering operations (bilinear, bicubic, lanczos) selectively to certain blocks based on their characteristics, rather than uniformly to all blocks. This partial application of excessive action achieves high prediction accuracy where needed while controlling overall encoding complexity.
2Measurement precision
If more neighboring samples are used for prediction, then the prediction accuracy improves, but the operation complexity increases
Solution Approach 1:
The patent uses more neighboring samples (including extended template regions) selectively for blocks that benefit from it, rather than applying it universally. This partial application improves prediction accuracy for complex regions while maintaining encoding efficiency for simpler regions.
Solution Approach 2:
Different prediction strategies are applied to different regions based on their local characteristics. Blocks with high frequency content or complex textures use extended neighboring samples and complex filters, while smooth blocks use simpler methods, optimizing the balance between accuracy and efficiency.
3Measurement precision
If complex filtering operations are applied, then the prediction sample accuracy improves, but the computational load increases
Solution Approach 1:
Complex filtering operations (bicubic, lanczos) are applied selectively to certain blocks based on their characteristics, rather than uniformly to all blocks. This partial application achieves high prediction sample accuracy where needed while controlling overall filtering complexity.
Solution Approach 2:
The patent changes filtering parameters (filter type, kernel size) based on block characteristics such as gradient magnitude and variance. This adaptive parameter selection achieves high prediction accuracy for complex blocks while using simpler filters for smooth blocks, managing overall computational load.
4Reliability
If adaptive filter selection is used, then the coding performance improves, but the decision complexity increases
Solution Approach 1:
The patent changes filter parameters (type, strength) based on block characteristics such as gradient magnitude and variance. This adaptive parameter selection improves coding performance by matching filter characteristics to block content while managing decision complexity through a limited set of predefined filter options.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An image decoding method and device according to the present disclosure may determine a reference block for intra prediction of a current block, generate a prediction sample of the current block by applying a filter to at least one of a reference sample belonging to the reference block, at least one neighboring sample adjacent to the reference sample, and a predetermined offset, and restore the current block on the basis of the prediction sample.