Selective Pre-Filtering for Video Coding Artifact Reduction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Traditional video codecs, such as HEVC, face challenges in efficiently reducing coding artifacts like ringing and banding artifacts, particularly in textured areas and uniform regions, due to limitations in existing in-loop filtering techniques.

Innovation Solution

The method involves pre-filtering pixels in reconstructed image units before applying offset values, using techniques like median filtering, to selectively address and reduce coding artifacts by preserving edges and minimizing noise.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional in-loop filtering (SAO) is applied to reduce coding artifacts, then coding artifacts like ringing and banding are reduced, but the filtering cannot selectively address different artifact types in different regions

Engineering Contradiction:
Improveartifact reduction effectivenessVSAvoidselective filtering capability
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent divides the image into different regions based on texture characteristics (flat regions vs. textured regions) and applies different filtering strategies to each region. Flat regions receive strong filtering to remove banding artifacts, while textured regions receive minimal or no filtering to preserve detail and avoid introducing new artifacts.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies different filtering strengths and types to different local regions of the image based on their characteristics. The filtering is adapted locally rather than uniformly applied, allowing optimal artifact reduction in each region without compromising overall image quality.

Inventive Principle:
Principle #3Local quality

2Object-affected harmful factors

If strong filtering is applied to remove banding artifacts in uniform regions, then banding artifacts are reduced, but ringing artifacts may be introduced or worsened in textured areas

Engineering Contradiction:
Improvebanding artifact reductionVSAvoidringing artifact introduction
Core Design Contradiction:
Object-affected harmful factorsVSObject-generated harmful factors

Solution Approach 1:

The patent applies different filtering strengths to different local regions based on their texture characteristics. Flat regions receive strong filtering to remove banding, while textured regions receive minimal filtering to avoid creating ringing artifacts, thus locally optimizing the filtering effect for each region type.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent uses the presence of texture as a indicator to adjust filtering strength, converting what would normally be a problem (texture making filtering difficult) into a benefit (texture guiding the filtering decision). By detecting texture, the system avoids over-filtering and the associated ringing artifacts.

Inventive Principle:
Principle #22Blessing in disguise (Convert harm into benefit)

3Reliability

If offset values are applied to all pixels to reduce coding artifacts, then overall image quality improves, but computational complexity and processing time increase

Engineering Contradiction:
Improveoverall image qualityVSAvoidencoding/decoding speed
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent applies filtering selectively only where needed (in flat regions) rather than uniformly to all pixels. This partial action approach maintains image quality in regions that benefit from filtering while avoiding unnecessary processing in regions where filtering would be harmful or ineffective, thus improving processing efficiency.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP3220644B1Method and device for encoding at least one image unit, and method and device for decoding a stream representative of at least one image unit and a computer program product
Publication Date: 2022.08.10 INTERDIGITAL VC HOLDINGS INC
  • EP3220644B1 patent drawingFigure 1
  • EP3220644B1 patent drawingFigure 2~8
  • EP3220644B1 patent drawingFigure 3

AI summary

A method for encoding at least one image unit is disclosed that comprises: - encoding (41) said at least one image unit, - reconstructing (42) said at least one image unit into a reconstructed image unit, - applying (43) a sample adaptive offset filter to said at least one reconstructed image unit ; wherein applying a sample adaptive offset filter to said at least one reconstructed image unit comprises: - determining (431) at least one current category in a set of categories for at least one pixel of the reconstructed image unit, - encoding (432) an offset value for at least one category of said set of categories ; - applying (433) a pre-filter to the at least one pixel of the reconstructed image unit assigned to said category only in the case where said current category is equal to said category having an encoded offset value; - adding (434) said encoded offset value to said at least one pixel of the reconstructed image unit.