Noise Suppression Filter Positioning in Video Coding Loop

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current video compression techniques face challenges in achieving improved compression ratios with minimal sacrifice in image quality, particularly due to issues like blocking artefacts and quantization noise, which can be exacerbated by the global application of noise suppression filters that increase coding complexity and file size.

Innovation Solution

The method involves strategically applying a noise suppression filter in various positions within the in-loop filtering process, such as before or after deblocking, sample adaptive offset, and adaptive loop filters, and controlling its application through flags indicating sequence, slice, or block-level application, to optimize compression efficiency and image quality.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Object-affected harmful factors

If a noise suppression filter is applied globally to the reconstructed image, then quantization noise and blocking artifacts are reduced, but coding complexity and file size increase

Engineering Contradiction:
Improvequantization noise and blocking artifactsVSAvoidcoding complexity
Core Design Contradiction:
Object-affected harmful factorsVSDevice complexity

Solution Approach 1:

The patent applies noise suppression filtering selectively at different granularities (sequence level, slice level, or block level) rather than globally. The filter is applied only to specific regions where blocking artifacts are detected, using flags to control application scope. This localizes the filtering operation to areas needing improvement, reducing overall coding complexity while maintaining effective noise suppression where required.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent segments the reconstructed image into different regions (sequences, slices, or blocks) and applies noise suppression filtering independently to each segment based on local characteristics. By dividing the image and applying filters selectively to segments with high artifact content, the system reduces global coding complexity while maintaining targeted noise suppression effectiveness.

Inventive Principle:
Principle #1Segmentation

2Object-affected harmful factors

If a noise suppression filter is applied globally to the reconstructed image, then quantization noise and blocking artifacts are reduced, but file size increases

Engineering Contradiction:
Improvequantization noise and blocking artifactsVSAvoidfile size
Core Design Contradiction:
Object-affected harmful factorsVSQuantity of substance

Solution Approach 1:

The patent reduces file size by applying noise suppression filtering only to specific local regions (sequences, slices, or blocks) where blocking artifacts are present, rather than globally processing the entire image. This selective application minimizes the additional data generated by filtering operations while maintaining effective noise suppression in critical areas.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent segments the image and applies filtering only to segments requiring improvement, reducing the total quantity of filtered data and consequently reducing the increase in file size. By dividing the image into sequences, slices, or blocks and selectively filtering only those with high artifact content, the system minimizes file size expansion while maintaining noise suppression effectiveness.

Inventive Principle:
Principle #1Segmentation

3Productivity

If noise suppression filter application is controlled at sequence level, then processing efficiency is improved, but image quality consistency across different regions deteriorates

Engineering Contradiction:
Improveprocessing efficiencyVSAvoidimage quality consistency
Core Design Contradiction:
ProductivityVSManufacturing precision

Solution Approach 1:

The patent introduces dynamic control mechanisms that allow the noise suppression filter application level to be adjusted based on image characteristics and processing requirements. The system can dynamically switch between sequence-level, slice-level, or block-level filtering, enabling adaptation to different content types and quality requirements, thus balancing processing efficiency with image quality consistency.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent applies noise suppression filtering at different granularities (sequence, slice, or block level) depending on the specific requirements. By allowing partial application (sequence level for efficiency) or more granular application (block level for quality consistency), the system can adjust the extent of filtering action to balance processing efficiency with image quality consistency across different regions.

Inventive Principle:
Principle #16Partial or excessive action

4Manufacturing precision

If noise suppression filter is applied at block level, then image quality consistency is improved, but coding complexity increases

Engineering Contradiction:
Improveimage quality consistencyVSAvoidcoding complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent applies noise suppression filtering at block level only where necessary, using flags to control application. By applying partial filtering (only to blocks with high artifact content) rather than excessive filtering (to all blocks), the system achieves improved image quality consistency in critical regions while minimizing the increase in coding complexity.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP3566454B1Noise suppression filter
Publication Date: 2024.07.31 HUAWEI TECH CO LTD
  • EP3566454B1 patent drawingFigure 1
  • EP3566454B1 patent drawingFigure 2
  • EP3566454B1 patent drawingFigure 3

AI summary

A mechanism of video coding is provided. The mechanism includes generating a reconstructed image from an encoded video stream. The reconstructed image is filtered to create a filtered image. The filtering includes applying a noise suppression filter to the reconstructed image. The noise suppression filter may be applied immediately prior to applying a deblocking filter to the reconstructed image, between a deblocking filter and a sample adaptive offset (SAO) filter, between the SAO filter and an adaptive loop filter, or after the adaptive loop filter. The filtered image is then stored in a picture buffer in memory for use in encoding or for output to a display in decoding.