Image Correction Using Region-Specific Weight Coefficients

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional image correction methods often result in incoordination by uniformly correcting the entire image, failing to consider the specificity of target regions, such as faces, which can lead to undesirable changes in the background.

Innovation Solution

The method sets different image weight coefficients for various regions within an image, allowing for tailored correction degrees to avoid incoordination by selectively correcting only the target regions, such as faces, while maintaining the integrity of the background.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If full-image correction is applied uniformly across the entire image, then image distortion is reduced, but incoordination occurs in the background and target region harmony deteriorates

Engineering Contradiction:
Improveimage distortion correctionVSAvoidbackground coordination
Core Design Contradiction:
Manufacturing precisionVSStability of the object's composition

Solution Approach 1:

The patent applies different correction strengths to different regions by introducing a weight coefficient map where each pixel has a specific weight value. The target region (e.g., face) has higher weight coefficients for stronger correction, while the background has lower weight coefficients to maintain its original appearance and avoid incoordination. This resolves the contradiction by making correction quality spatially variable rather than uniform.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent segments the image into different regions with different correction requirements by generating a weight coefficient map that divides the image into target regions (requiring strong correction) and background regions (requiring weak or no correction). This segmentation allows selective application of correction strength to different parts of the image, preventing background incoordination while maintaining target region correction quality.

Inventive Principle:
Principle #1Segmentation

2Manufacturing precision

If strong correction is applied to the target region, then distortion in the target region is reduced, but the background may become distorted and incoordinate

Engineering Contradiction:
Improvetarget region distortion correctionVSAvoidbackground distortion
Core Design Contradiction:
Manufacturing precisionVSObject-affected harmful factors

Solution Approach 1:

The patent uses a weight coefficient map to assign different correction intensities to different spatial locations. The target region is assigned high weight coefficients (close to 1.0) to receive strong correction, while the background is assigned low weight coefficients (close to 0.0) to receive minimal or no correction. This local differentiation eliminates the harmful effect of background distortion while maintaining effective target region correction.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent applies partial correction action by using the weight coefficient map to selectively apply correction only where needed. Instead of applying full correction uniformly, the system applies correction proportionally based on the weight coefficients, resulting in partial correction for background regions and full correction for target regions, thus avoiding excessive correction in the background.

Inventive Principle:
Principle #16Partial or excessive action

3Ease of operation

If uniform correction is applied to the entire image, then processing simplicity is maintained, but region-specific correction needs are not met

Engineering Contradiction:
Improvecorrection process simplicityVSAvoidregion-specific correction
Core Design Contradiction:
Ease of operationVSAdaptability or versatility

Solution Approach 1:

The patent implements self-service by automatically generating the weight coefficient map through algorithms that identify target regions (such as face detection and analysis) and compute appropriate weight coefficients for each pixel. This automated approach maintains operational simplicity while achieving region-specific correction, as the system performs the complex region identification and weight assignment tasks autonomously without requiring manual intervention.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent changes the correction parameter from a uniform scalar value to a spatially varying weight coefficient field. By representing correction strength as a function of position (through the weight coefficient map), the system achieves region-specific adaptability while maintaining a unified correction framework. This parameter transformation allows the system to handle complex region-specific requirements without fundamentally changing the correction process architecture.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3751508B1Method and device for image correction and storage medium
Publication Date: 2023.05.31 BEIJING XIAOMI MOBILE SOFTWARE CO LTD
  • EP3751508B1 patent drawingFigure 1A~1B
  • EP3751508B1 patent drawingFigure 2
  • EP3751508B1 patent drawingFigure 3

AI summary

The present disclosure relates to a method and device for image correction and a storage medium. The method includes: a correction offset for each unit to be corrected in an image is determined; at least one target region in the image is determined; an image weight coefficient for each unit to be corrected in the image is determined according to the target region; a final offset for each unit to be corrected in the image is determined according to the image weight coefficient and the correction offset; and each unit to be corrected in the image is corrected according to the final offset. Through the technical solutions of the present disclosure, the image weight coefficients are set, so that different regions in the image may be corrected to different degrees as required.