Defect Detection in Printed Documents via Activity Matrix Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting defects in printed documents, such as image streaking and graininess, struggle with small differences between reference and target images, color variations, and misalignment, leading to ineffective defect detection.
Innovation Solution
A system that analyzes activity matrices for both reference and target images using low-pass filters to classify pixels as defective, employing morphological operations and threshold comparisons to identify defects, and combines results from swapped image analyses to enhance detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If global comparison models (SIFT, histogram comparisons) or pixel-by-pixel local differences (SSIM) are used to detect defects, then the system can identify obvious differences between reference and target images, but the system fails to detect defects where differences are small, colors vary, or alignment is imperfect
Solution Approach 1:
The patent transforms the defect detection approach by changing from direct image comparison to activity matrix comparison. It applies low-pass filtering to convert images into activity matrices that represent local variations, then compares these transformed representations using a threshold on activity differences. This parameter transformation enables detection of subtle defects while being invariant to color variations and alignment issues.
Solution Approach 2:
The patent introduces activity matrices as an intermediary representation between the reference and target images. Instead of directly comparing images or pixels, the system first converts both images into activity matrices through low-pass filtering, then performs comparison on these intermediate representations. This intermediary step decouples the detection process from color and alignment variations.
2Measurement precision
If gray level comparison between reference and target images is performed, then the system can detect differences in printed documents, but the method does not perform well due to difficulty in accurately matching colors
Solution Approach 1:
The patent extracts only the relevant information for defect detection by converting images into activity matrices that capture local variations while discarding absolute color information. The low-pass filtering process extracts activity patterns that are invariant to color variations, allowing defect detection without complex color matching procedures.
3Reliability
If the system analyzes activity matrices and compares pixel activity levels with thresholding, then the system can detect defects with small differences and color variations, but the system requires morphological operations and multiple processing steps
Solution Approach 1:
The patent segments the defect detection process into distinct stages: (1) low-pass filtering to create activity matrices, (2) activity difference calculation, (3) threshold-based defect classification, and (4) morphological operations for refinement. This segmentation allows each stage to focus on a specific aspect, improving overall reliability while making the complex process more manageable and systematic.
Data Source
AI summary
An image system for detecting defects in an image may include a processing device to detect defects in a target image by analyzing the activity level in a reference and the target image. The system may identify a quiet area/pixel in the reference image based on the activity level, check the activity level of the corresponding pixel in the target image, and classify the pixel in the target image as defective if the activity level of the pixel in the target image exceeds a noise threshold. The system may additionally swap the reference and target image, repeat the detection steps and combine the detection results with those before the swap. The system may also include an image sensor to scan a printed document of the reference image into a scanned electronic document (the target image).