Template Matching Using Correlation Maps for Distorted Patterns

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing image inspection techniques, particularly in semiconductor wafer measurement, face challenges in achieving accurate template matching due to distorted pattern shapes and variations in edge intensity, which can lead to misalignment and noise emphasis issues.

Innovation Solution

The proposed solution involves generating a correlation value contribution rate map to identify key areas contributing to similarity between template and search images, followed by luminance correction and local matching to refine candidate positions, and optionally using a centroid distance filter to tolerate distortion without edge expansion.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If template matching is performed on distorted patterns without distortion tolerance, then measurement precision is improved, but matching reliability deteriorates due to inability to recognize distorted patterns as matches

Engineering Contradiction:

Improveposition determination accuracyVSAvoidpattern recognition reliability

Core Design Contradiction:

Measurement precisionVSReliability

Solution Approach 1:

The patent changes the parameter of edge thickness in the correlation value contribution rate map to tolerate distortion. By creating a map where correlation values are distributed with thicker edges compared to the original edge image, the system can recognize distorted patterns as matches while maintaining accurate position determination.

Inventive Principle:

Principle #35Parameter changes

2Difficulty of detecting and measuring

If luminance correction is performed to emphasize low-intensity edges, then detection of low-intensity edges is improved, but noise emphasis occurs leading to false matching

Engineering Contradiction:

Improvelow-intensity edge detectionVSAvoidnoise emphasis

Core Design Contradiction:

Difficulty of detecting and measuringVSObject-generated harmful factors

Solution Approach 1:

The patent applies local quality by performing luminance correction selectively only in regions where the correlation value contribution rate exceeds a predetermined threshold. This ensures that low-intensity edges are emphasized only in relevant areas, while noise regions with low correlation values remain unaffected, preventing false matching.

Inventive Principle:

Principle #3Local quality

3Adaptability or versatility

If edge images are thickened to tolerate distortion, then adaptability to distorted patterns is improved, but measurement precision deteriorates due to loss of precise edge position information

Engineering Contradiction:

Improvedistortion toleranceVSAvoidedge position accuracy

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent segments the image processing into two distinct stages: first, edge detection to obtain precise edge positions; second, correlation value contribution rate map creation with edge thickening for distortion tolerance. This segmentation allows each stage to optimize for its specific function without compromising the other.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary edge detection and position identification before creating the correlation value contribution rate map with thickened edges. This preliminary action ensures that precise edge position information is captured first, and the subsequent thickening is used only for tolerance without losing the original precision data.

Inventive Principle:

Principle #10Preliminary action

4Productivity

If template matching is performed on original edge images without correlation value contribution rate map, then processing speed is improved, but matching accuracy deteriorates due to inability to handle distorted patterns and luminance variations

Engineering Contradiction:

Improveprocessing speedVSAvoidmatching accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent implements partial action by creating the correlation value contribution rate map only for regions exceeding a predetermined correlation threshold, rather than processing the entire image. This partial processing maintains reasonable speed while improving accuracy for relevant regions.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS8498489B2Image inspection apparatus

Publication Date: 2013.07.30 HITACHI HIGH TECH CORP

AI summary

The present invention provides a unified template matching technique which allows an adequate matching position to be provided even in an image with a distorted pattern shape and a variation in edge intensity. A correlation value contribution rate map is created for the vicinity of each of top candidate positions obtained by applying a centroid distance filter to a normalized correlation map resulting from template matching. A corrected intensity image is created from the correlation value contribution rate maps. Luminance correction is performed based on the corrected intensity image. Local matching is performed again on the vicinity of each candidate position. The candidates are then re-sorted based on candidate positions and correlation values newly obtained. Thus, even in an image with a distorted pattern shape and a variation in edge intensity, an adequate matching position can be provided in a unified manner.