Image Correction Device Shadow Region Estimation Clustering

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing technologies face challenges in accurately detecting shadow regions due to the inability to directly utilize shadow boundary information, leading to potential detection omissions and unnatural boundary corrections, especially with instruments having large reflection intensity errors.

Innovation Solution

An image correction device and method that includes an input processing unit for receiving images and three-dimensional point clouds with reflection intensity, a shadow region estimation unit that performs clustering based on pixel values and positions to estimate shadow regions, and a shadow correction unit that corrects pixel values in the estimated shadow regions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Difficulty of detecting and measuring

If shadow boundary detection is performed using gradient information on reflection intensity and color gradient, then shadow detection capability is improved, but detection accuracy deteriorates due to inability to directly use shadow boundary information

Engineering Contradiction:

Improveshadow detection capabilityVSAvoidshadow detection accuracy

Core Design Contradiction:

Difficulty of detecting and measuringVSMeasurement precision

Solution Approach 1:

The patent segments the image into multiple clusters based on pixel values and positions, then performs shadow detection on each cluster separately. This segmentation allows the use of shadow boundary information within each cluster while maintaining overall detection capability, resolving the contradiction between detection capability and accuracy.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces a new dimension by combining reflection intensity information with color information in a unified shadow detection framework. By using both gradient information on reflection intensity and color gradient simultaneously, the system achieves accurate shadow boundary detection that directly utilizes shadow boundary information rather than relying on indirect methods.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

2Measurement precision

If clustering is performed on pixels based on pixel values and positions, then shadow region estimation accuracy is improved, but device complexity increases

Engineering Contradiction:

Improveshadow region estimation accuracyVSAvoidprocessing complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent divides the image into multiple clusters based on pixel values and positions, enabling accurate shadow region estimation within each cluster. This segmentation approach improves estimation accuracy by localizing shadow detection to specific regions while managing complexity through modular processing of individual clusters.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent performs clustering and shadow detection on selected regions (clusters) rather than the entire image uniformly. By applying shadow detection selectively to clustered regions where shadows are likely to occur, the system achieves high accuracy while reducing overall processing complexity compared to full-image processing.

Inventive Principle:

Principle #16Partial or excessive action

3Area of stationary object

If reflection intensity measurement is performed with instruments having large error, then measurement coverage is improved, but measurement precision deteriorates

Engineering Contradiction:

Improvemeasurement coverageVSAvoidreflection intensity precision

Core Design Contradiction:

Area of stationary objectVSMeasurement precision

Solution Approach 1:

The patent merges reflection intensity information with color information in a unified shadow detection framework. By combining multiple types of data (reflection intensity gradients and color gradients), the system compensates for errors in reflection intensity measurements while maintaining broad measurement coverage, achieving both coverage and precision.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent uses clustering results and shadow boundary detection feedback to refine reflection intensity measurements. By iteratively adjusting measurements based on shadow boundary information and cluster analysis, the system corrects errors in reflection intensity data while maintaining comprehensive measurement coverage.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS20250045875A1Image correction device, image correction method, and image correction program

Publication Date: 2025.02.06 NIPPON TELEGRAPH & TELEPHONE CORP

AI summary

An input processing unit receives an image and a three-dimensional point cloud including three-dimensional points having reflection intensity on a surface of an object for which at least a relationship between an image capturing position and a measurement position is obtained in advance, and obtains pixel positions on the image corresponding to the respective three-dimensional points of the three-dimensional point cloud. A shadow region estimation unit performs clustering on pixels of the image on the basis of pixel values and pixel positions, obtains an average reflection intensity and an average value of quantified color information for each of clusters, and estimates a shadow region by performing comparison in the average reflection intensity and the average value of the color information between the clusters. A shadow correction unit corrects pixel values of the shadow region from the shadow region estimated and the image.