Inter Prediction Filtering for Motion-Based Block Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image compression technologies face challenges in achieving high efficiency for compressing high-resolution and high-quality images, particularly in accurately predicting pixel values for inter prediction.
Innovation Solution
An image decoding method that determines a reference block based on motion information, generates a prediction sample by applying a filter to reference and neighboring samples, and reconstructs the current block using these samples, with filter coefficients derived from specific neighboring areas and templates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a filter is applied to reference samples and neighboring samples to generate prediction samples, then prediction accuracy is improved, but operation complexity increases
Solution Approach 1:
The patent applies different filtering operations to different neighboring areas (first neighboring area and second neighboring area) based on their specific characteristics. The filter coefficients are derived separately for each area, allowing optimized prediction for each local region while managing overall complexity through structured processing.
Solution Approach 2:
The patent divides the neighboring areas into multiple regions (first neighboring area and second neighboring area) and processes them separately. This segmentation allows the system to derive filter coefficients from specific local regions and apply them appropriately, balancing prediction accuracy with operational complexity.
2Productivity
If filter coefficients are derived from multiple neighboring areas and templates, then encoding performance is improved, but processing time increases
Solution Approach 1:
The patent performs preliminary actions by deriving filter coefficients from neighboring areas and templates before the actual prediction process. These pre-computed filter coefficients can then be efficiently applied during encoding, reducing the time required for prediction while maintaining high encoding performance.
Solution Approach 2:
The patent uses templates that copy and store characteristic patterns from neighboring areas. These templates serve as reusable references that can be quickly applied during prediction, reducing the need for complex real-time calculations and thereby decreasing processing time while maintaining encoding performance.
3Measurement precision
If multiple template candidates are used to represent block characteristics, then prediction accuracy is improved, but device complexity increases
Solution Approach 1:
The patent creates multiple template candidates, each representing different local characteristics of neighboring areas. By selecting appropriate templates based on local block characteristics, the system achieves high prediction accuracy without requiring all possible templates to be processed simultaneously, thus managing device complexity.
Solution Approach 2:
The patent dynamically selects from multiple template candidates based on the specific characteristics of the current block and neighboring areas. This dynamic selection process allows the system to adapt to different image content while maintaining manageable complexity by only processing relevant templates for each specific case.
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 of the current block on the basis of motion information about the current block; generate prediction samples of the current block by applying a filter to at least one of a reference sample belonging to the reference block, one or more neighboring samples adjacent to the reference sample, or a prescribed offset; and reconstruct the current block on the basis of the prediction samples.