Adaptive Intra Prediction Filtering for Video Block Coding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video signal processing methods require more efficient coding techniques to enhance coding efficiency and signaling efficiency related to motion information sets.

Innovation Solution

A video signal processing method that involves deriving an intra-prediction mode, configuring reference samples, generating prediction samples using interpolation filters based on filter flags, and reconstructing blocks with specific filter coefficients, and includes intra-sub-partition modes for splitting blocks into rectangular transform blocks.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If conventional video compression techniques are used, then basic compression is achieved, but coding efficiency is insufficient

Engineering Contradiction:
Improvecoding efficiencyVSAvoidvideo signal quality
Core Design Contradiction:
ProductivityVSLoss of information

Solution Approach 1:

The patent applies parameter changes by dynamically selecting transform kernels based on prediction mode and block characteristics. Different transform kernels (e.g., DCT, DST, KST) are chosen to optimize the transformation of residual signals, improving coding efficiency while maintaining video quality through adaptive parameter selection rather than fixed conventional transforms

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If more detailed motion information is signaled, then prediction accuracy improves, but signaling overhead increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidbitrate overhead
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

The patent segments the prediction process by dividing blocks into sub-blocks and applying different transform kernels to different regions based on local characteristics. This segmentation allows precise adaptation to local video content without signaling detailed motion information for every sub-region, reducing overhead while maintaining prediction accuracy

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces dynamic selection of transform kernels based on prediction mode and block characteristics. Instead of static conventional transforms, the system dynamically adapts the transformation process to match local video content, improving prediction accuracy without requiring additional signaling for motion parameters

Inventive Principle:
Principle #15Dynamics

3Ease of manufacture

If a single transform kernel is used for all blocks, then processing is simple, but transform suitability for specific blocks is poor

Engineering Contradiction:
Improveprocessing simplicityVSAvoidtransform efficiency
Core Design Contradiction:
Ease of manufactureVSProductivity

Solution Approach 1:

The patent applies local quality by selecting different transform kernels for different blocks based on their specific characteristics such as prediction mode and content type. Instead of uniform processing, each block receives an optimized transform kernel (e.g., DST for directional predictions, DCT for others), improving transform efficiency while keeping the overall process manageable through systematic selection rules

Inventive Principle:
Principle #3Local quality

4Productivity

If intra-sub-partition modes are applied, then block transformation is optimized, but processing complexity increases

Engineering Contradiction:
Improveblock transformation efficiencyVSAvoidprocessing complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent segments blocks into sub-blocks for independent transform processing, allowing optimized kernel selection for each sub-region. This segmentation improves transformation efficiency by adapting to local characteristics while the segmentation follows systematic patterns that manage processing complexity through regular division schemes

Inventive Principle:
Principle #1Segmentation

Data Source

PatentUS12413710B2Video signal processing method and device using intra prediction filtering
Publication Date: 2025.09.09 SAMSUNG ELECTRONICS CO LTD
  • US12413710B2 patent drawing
  • US12413710B2 patent drawing
  • US12413710B2 patent drawing

AI summary

A video signal processing method includes: a step for deriving an intra prediction mode of a current block; a step for constructing a reference sample around the current block; a step for generating a prediction sample of the current block by using the reference sample on the basis of the intra prediction mode; and a step for restoring the current block on the basis of the prediction sample. The step for generating the prediction sample may include: a step for setting a filter flag value which specifies a filter coefficient of an interpolation filter applied to the reference sample on the basis of the width and height of the current block; and a step for performing filtering on the reference sample by using the interpolation filter having the filter coefficient specified by the filter flag.