Print Inspection Alignment Using Content-Matched Deformation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image alignment techniques in print inspection systems fail to accurately align images with local distortions, leading to erroneous inspections when patterns in the reference image are not present in the inspection target image, increasing the rate of incorrect defect detection.
Innovation Solution
An inspection apparatus and method that determines content coincidence between the inspection target image and reference image, controlling alignment based on this determination to prevent erroneous inspection by dividing images into regions, using linear and nonlinear transformations, and setting control points only where contents coincide.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If free-form deformation alignment is used to handle local distortions, then alignment of images with local magnification and positional displacement is improved, but erroneous inspection occurs when patterns in reference image and inspection target image do not completely coincide
Solution Approach 1:
The patent divides the image alignment process into two distinct phases: a first alignment process using linear transformation for global alignment, and a second alignment process using free-form deformation for local distortion correction. This segmentation allows each alignment method to be applied appropriately, preventing erroneous deformation when patterns do not coincide while still achieving accurate local alignment where patterns match.
Solution Approach 2:
The patent performs preliminary global alignment using linear transformation before executing the local alignment using free-form deformation. This preliminary action ensures that the images are roughly aligned first, providing a foundation for the subsequent local distortion correction and preventing erroneous deformation of non-matching patterns.
2Ease of manufacture
If alignment by linear transformation is used, then processing simplicity is maintained, but adequate alignment cannot be performed when local distortions are present in scanned images
Solution Approach 1:
The patent segments the alignment process into two distinct stages: first applying linear transformation for simple global alignment, then applying free-form deformation for complex local distortion correction. This segmentation maintains processing simplicity for the initial alignment while achieving high precision for local distortions through the subsequent detailed alignment process.
Solution Approach 2:
The patent employs a dynamic, multi-stage alignment approach where the alignment method is adapted based on the distortion characteristics. Linear transformation is used for the initial global alignment phase, and free-form deformation is dynamically applied for the second phase to handle local distortions, allowing the system to flexibly adjust to different distortion scenarios.
3Ease of operation
If free-form deformation is applied to images with non-coinciding patterns, then alignment processing is performed, but control points are erroneously optimized minimizing incorrect difference values, increasing erroneous inspection rate
Solution Approach 1:
The patent performs preliminary global alignment using linear transformation before applying free-form deformation. This preliminary action ensures that images are roughly aligned and patterns are approximately coincident before the second alignment process, preventing erroneous optimization of control points and reducing false defect detection.
Solution Approach 2:
The patent divides alignment into two phases: first linear transformation for global alignment, then free-form deformation for local alignment. This segmentation ensures that free-form deformation is only applied when images are already roughly aligned, preventing erroneous processing of non-coinciding patterns while maintaining the ability to handle local distortions.
Data Source
AI summary
The present invention provides an inspection apparatus operable to read an image formed on a printing medium by a printing apparatus and inspect the read image. The inspection apparatus stores image data used at the time of forming the image on a printing medium as a reference image, obtains image data of an image of a printing medium on which the reference image has been formed as a inspection target image and determines whether contents of the inspection target image and contents of the reference image coincide. The inspection apparatus aligns the inspection target image and the reference image according to a result of the determination; and detects a defect in the inspection target image by comparing the aligned inspection target image and reference image.