Sample Adaptive Offset Classification for Pixel-Value Variation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video compression methods, such as HEVC, are not well adapted to different variations in image pixel content and types of components, leading to inefficiencies in encoding and decoding processes.

Innovation Solution

Adaptive classification of pixels into classes based on their pixel values, using a rate distortion criterion to determine optimal classification groups, and applying sample adaptive offset (SAO) filtering to improve encoding and decoding efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If fixed classification ranges are used for SAO filtering, then the encoding process is simple, but it cannot adapt to different variations in image pixel content and component types

Engineering Contradiction:
Improveadaptation to pixel value distributionsVSAvoidencoding complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent applies dynamics by making the classification ranges adaptive rather than fixed. The encoder dynamically determines optimal classification ranges based on the actual pixel value distribution in the image data, allowing the SAO filtering to adapt to different variations in image content and component types while maintaining encoding efficiency

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent changes the parameter of classification ranges from fixed predetermined values to dynamically determined values based on pixel value distributions. This allows the system to optimize the classification parameters according to the specific characteristics of each image, improving adaptability without significantly increasing complexity

Inventive Principle:
Principle #35Parameter changes

2Manufacturing precision

If additional classification data is transmitted to improve adaptability, then visual quality improves, but bitrate increases

Engineering Contradiction:
Improvevisual qualityVSAvoidbitrate
Core Design Contradiction:
Manufacturing precisionVSQuantity of substance

Solution Approach 1:

The patent applies local quality by determining classification ranges specifically for each component type (luma and chroma) based on their respective pixel value distributions. This localized adaptation allows the system to improve visual quality for each component according to its specific characteristics without transmitting excessive general data

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The system uses the pixel value distribution information that is already available during encoding to automatically determine the optimal classification ranges. This self-service approach allows the encoder to adapt to different image contents without requiring additional side information to be transmitted, thus improving visual quality without increasing bitrate

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP3648461B1Method and device for providing compensation offsets for a set of reconstructed samples of an image
Publication Date: 2026.01.07 CANON KK
  • EP3648461B1 patent drawingFigure 1
  • EP3648461B1 patent drawingFigure 2
  • EP3648461B1 patent drawingFigure 3

AI summary

Compensation offsets are provided for a set of reconstructed samples of an image. Each sample has a sample value. A method of providing the compensation offsets comprises selecting, based on a rate distortion criterion, a classification from among a plurality of predetermined classifications. Each predetermined classification has a classification range smaller than a full range of the sample values and is made up of a plurality of classes, each defining a range of sample values within the classification range, into which class a sample is put if its sample value is within the range of the class concerned. A compensation offset is associated with each class of the selected classification for application to the sample value of each sample of the class.