Exposure Compensation via Masked Reference Image Averaging

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image and video capture systems face challenges in maintaining consistent exposure settings over time, leading to inconsistencies in image quality due to automatic adjustments, which can hinder the effectiveness of video processing algorithms when exposure settings are unknown or varying.

Innovation Solution

A method that generates a masked reference image by applying a mask to exclude changing objects and calculates exposure adjustment estimates by averaging pixel value differences between corresponding pixels in two images, allowing for compensation of exposure adjustments to ensure consistent image processing.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Illumination intensity

If automatic exposure adjustments are made between frames, then image quality is improved under varying lighting conditions, but consistency of exposure representation over time deteriorates

Engineering Contradiction:
Improveimage qualityVSAvoidexposure consistency
Core Design Contradiction:
Illumination intensityVSStability of the object's composition

Solution Approach 1:

The system performs preliminary actions by capturing a reference frame and calculating exposure adjustments before processing subsequent frames. The exposure adjustment is computed based on the reference frame and applied to compensate for lighting changes, ensuring consistent exposure representation across frames while allowing automatic exposure adjustments to improve image quality under varying conditions.

Inventive Principle:
Principle #10Preliminary action

2Adaptability or versatility

If digital exposure correction is applied dynamically, then adaptability to changing scenes is improved, but processing complexity increases

Engineering Contradiction:
Improveadaptability to changing scenesVSAvoidprocessing complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The processing is segmented into distinct stages: capturing a reference frame, calculating exposure adjustment based on the reference frame, and applying the adjustment to subsequent frames. This segmentation allows the system to adapt to changing scenes through dynamic exposure correction while managing processing complexity by breaking down the task into manageable, repetitive operations that can be efficiently implemented.

Inventive Principle:
Principle #1Segmentation

3Speed

If exposure settings vary between frames, then responsiveness to lighting changes is improved, but reliability of video processing algorithms deteriorates

Engineering Contradiction:
Improveresponsiveness to lighting changesVSAvoidalgorithm effectiveness
Core Design Contradiction:
SpeedVSReliability

Solution Approach 1:

The system uses feedback by continuously monitoring lighting conditions and comparing them to the reference frame, then adjusting exposure settings accordingly. This feedback mechanism allows the system to respond quickly to lighting changes while maintaining reliability of video processing algorithms by ensuring that exposure adjustments are systematically applied based on measured differences from the reference frame.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP3210377B1Compensation of difference in exposure between two images
Publication Date: 2020.12.02 MICROSOFT TECHNOLOGY LICENSING LLC
  • EP3210377B1 patent drawingFigure 1
  • EP3210377B1 patent drawingFigure 2
  • EP3210377B1 patent drawingFigure 3

AI summary

Unknown exposure differences between two images of the same scene are compensated based on the common image portion: an compensation value is estimated by averaging, from the pixels of the first image having a particular pixel value in the common image portion, the differences between said particular pixel value and value of correpsonding pixels in the second image. The images may be in any color space for visible light, or images not of visible light, such as infrared or depth images.