Extended Taps for Adaptive Loop Filter in Video Coding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current video coding technologies, such as those in HEVC and VVC, face limitations in utilizing a wide range of information for adaptive loop filter (ALF) training and filtering, often relying solely on spatial reconstruction samples from the current frame, neglecting valuable information from reference frames and other processing stages like deblocking filters and transform domains.

Innovation Solution

The introduction of extended taps in ALF filters that can utilize information from spatial and temporal neighbors, including reference frames and different processing stages, to enhance filtering efficiency by incorporating multiple inputs and adaptive filter shapes and lengths.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If only spatial reconstruction samples from the current frame are used for ALF filtering, then the filtering process is simple and fast, but the filtering effectiveness and artifact reduction capability are limited

Engineering Contradiction:
Improvefiltering effectivenessVSAvoidfiltering process complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent extends the filtering approach from a single spatial dimension (current frame only) to multiple dimensions by incorporating temporal information from reference frames. The extended tap structure accesses samples from both spatial neighbors and temporal reference frames, creating a multi-dimensional filtering approach that improves artifact reduction while maintaining manageable complexity through structured access patterns.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Solution Approach 2:

The patent combines multiple types of samples (spatial reconstruction samples from current frame and temporal samples from reference frames) into a composite filtering input. This composite approach integrates diverse information sources to enhance filtering effectiveness, similar to how composite materials combine different substances to achieve superior properties.

Inventive Principle:
Principle #40Composite materials

2Productivity

If extended taps utilizing multiple inputs are used in ALF, then coding efficiency and artifact reduction improve, but computational complexity and processing overhead increase

Engineering Contradiction:
Improvecoding efficiencyVSAvoidcomputational complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent segments the filtering process into distinct components: spatial taps for current frame samples and extended taps for reference frame samples. This segmentation allows the system to selectively apply different filtering approaches to different regions and components, improving coding efficiency through targeted filtering while managing computational complexity by avoiding uniform complex processing across all blocks.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements adaptive filtering where the tap configuration (spatial vs. extended) dynamically adjusts based on block characteristics, motion information, and regional needs. This dynamic adaptation allows the system to use simpler spatial taps when sufficient and more complex extended taps only when beneficial, optimizing the balance between coding efficiency and computational complexity.

Inventive Principle:
Principle #15Dynamics

3Adaptability or versatility

If ALF uses fixed filter shapes and lengths, then implementation is simpler and faster, but adaptability to different regions and components is reduced

Engineering Contradiction:
Improveadaptability to regions and componentsVSAvoidfilter configuration complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent applies different filter configurations (spatial taps vs. extended taps, different shapes and lengths) to different regions and components based on local characteristics. Luma and chroma components receive tailored filtering, and different blocks within the picture can use different tap configurations matched to their specific needs, enhancing adaptability while maintaining implementation feasibility through structured classification.

Inventive Principle:
Principle #3Local quality

Data Source

PatentUS20250106394A1Extended taps using different sources for adaptive loop filter in video coding
Publication Date: 2025.03.27 BYTEDANCE INC
  • US20250106394A1 patent drawing
  • US20250106394A1 patent drawing
  • US20250106394A1 patent drawing

AI summary

A method for processing video data includes using one or more extended taps in an ALF or a CCALF and performing a conversion between a visual media data and a bitstream based on the extended taps used in the ALF. A syntax element structure in the bitstream contains one or more filters with at least one extended tap. A first syntax element is in the bitstream to indicate whether a filter with at least one extended tap is enabled.