Parameter-Based Texture Normalization for Multi-View Image Consistency

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for normalizing texture information from multiple camera sources with varying illumination conditions and viewpoints are inadequate, particularly when camera control is uncontrollable and handles uncontrollable camera parameters and artifacts like specularity and glare.

Innovation Solution

A method involving parameter-based transformation maps is used to optimize texture normalization across multiple images by detecting matching regions, identifying interdependencies, and calibrating 2D and 3D information to generate transformation maps that account for both matching and non-matching areas, ensuring visual consistency and handling uncontrollable camera parameters.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If color calibration with reference pattern is used, then color consistency is improved, but uncontrollable camera parameters and viewpoint-specific artifacts cannot be handled

Engineering Contradiction:

Improvecolor consistencyVSAvoidhandling of uncontrollable camera parameters

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent transforms the color calibration problem from adjusting camera parameters to optimizing transformation parameters that map colors between source and target images. Instead of controlling camera parameters (which are uncontrollable), the system learns optimal transformation parameters through iterative optimization that adapts to different viewing conditions and handles specularity automatically.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent creates a target image that represents the desired appearance, and then generates transformation maps that copy and adapt texture information from multiple source images to match this target. The transformation maps effectively copy the desired color properties while adapting to handle viewpoint-specific artifacts and specularity.

Inventive Principle:

Principle #26Copying

2Ease of manufacture

If histogram matching is used, then color transformation is simplified, but good results are only obtained when images show similar color statistics

Engineering Contradiction:

Improvecolor transformation processVSAvoidhandling of varying color statistics

Core Design Contradiction:

Ease of manufactureVSAdaptability or versatility

Solution Approach 1:

The patent replaces static histogram matching with a dynamic iterative optimization process. Instead of assuming similar color statistics and applying a fixed transformation, the system dynamically adjusts transformation parameters through multiple iterations, adapting to varying color statistics in different images and achieving consistent results even when source images have very different color distributions.

Inventive Principle:

Principle #15Dynamics

3Area of stationary object

If multiple cameras with different viewpoints are used, then coverage is improved, but viewpoint-specific artifacts like specularity and glare vary across images

Engineering Contradiction:

ImprovecoverageVSAvoidvisual consistency

Core Design Contradiction:

Area of stationary objectVSReliability

Solution Approach 1:

The patent segments the image processing problem by generating separate transformation maps for different regions of the object. The optimization process identifies matching regions across multiple viewpoint images and creates region-specific transformation parameters, allowing viewpoint-specific artifacts like specularity to be handled appropriately for each surface orientation while maintaining overall visual consistency.

Inventive Principle:

Principle #1Segmentation

4Measurement precision

If automatic camera adjustments are enabled, then image quality is improved, but texture normalization becomes unpredictable

Engineering Contradiction:

Improveimage qualityVSAvoidtexture normalization consistency

Core Design Contradiction:

Measurement precisionVSReliability

Solution Approach 1:

The patent inverts the traditional approach by not trying to control or disable automatic camera adjustments, but instead accepting them and compensating for their effects through learned transformation maps. The system works backwards from the actual captured images (with all automatic adjustments applied) to create transformations that normalize the texture, making the process reliable regardless of what automatic adjustments the cameras apply.

Inventive Principle:

Principle #13The other way round (Inversion)

Data Source

PatentUS10692269B2Method and device for image processing

Publication Date: 2020.06.23 ALCATEL LUCENT SA

AI summary

A method for generating a set of respective transformation maps for a set of respective 2D images from a same object and using a parameter-based transformation model, comprises the steps of —receiving said set of respective 2D images and said parameter-based transformation model —detecting matching regions across several pairs of the 2D images, based on the set of 2D images and 3D information of said object, —identifying respective interdependencies of the matching regions over the 2D images, —optimizing the parameters of the parameter-based transformation model over the matching regions of all images as well as over the non-matching regions in all images.