Line Mura Defect Detection in Displays Using Preprocessing and ML
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automated inspection techniques are ineffective in detecting Mura defects in display surfaces due to their low contrast and lack of distinct edges, making it difficult to identify defects like line Mura, spot Mura, and region Mura, especially in images with uneven illumination.
Innovation Solution
A system and method using a machine learning classifier to identify instances of horizontal line Mura by preprocessing images to reduce noise, dividing them into patches, extracting statistical features, and classifying using a supervised learning model trained with human input to enhance detection accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If automated surface inspection is used to detect defects, then detection speed increases, but detection accuracy deteriorates for low-contrast defects like Mura
Solution Approach 1:
The patent applies preprocessing operations (noise reduction filtering, illumination correction) before defect detection to enhance the visibility of low-contrast Mura defects. This preliminary enhancement of image quality enables automated inspection systems to accurately detect subtle brightness non-uniformities that would otherwise be invisible, resolving the contradiction between fast automated detection and accurate defect identification
2Measurement precision
If manual defect detection is used, then detection accuracy improves, but processing time increases excessively
Solution Approach 1:
The system performs preliminary image enhancement including noise reduction and illumination correction before classification. This preprocessing step automatically prepares the image data in a manner that would otherwise require manual inspection, enabling automated systems to achieve accuracy comparable to manual detection while maintaining high processing speed
Solution Approach 2:
The patent introduces an intermediary classification model that acts as a bridge between automated inspection and manual detection. The model is trained to recognize Mura defect patterns and classifies image regions automatically, providing expert-level detection accuracy without the time cost of manual inspection
3Reliability
If the entire image is classified without preprocessing, then comprehensive coverage is achieved, but processing time and computational resources increase
Solution Approach 1:
The patent divides the display image into multiple overlapping patches and processes each patch independently through the classification model. This segmentation approach enables parallel processing of image regions, reducing overall processing time while maintaining comprehensive defect coverage through the overlapping patch strategy that ensures no defect is missed at patch boundaries
4Measurement precision
If classification is performed on all image patches, then detection thoroughness improves, but computational complexity increases
Solution Approach 1:
The system applies preprocessing operations (noise reduction, illumination correction) to all image patches to enhance defect visibility, then uses the classification model to identify Mura defects. The preprocessing step reduces computational complexity by enhancing defect characteristics before classification, making the subsequent classification process more efficient while maintaining thorough detection
Data Source
Figure 1A
Figure 1B
Figure 2
AI summary
A system and method for identifying line Mura defects on a display. The system is configured to generate a filtered image by preprocessing an input image of a display using at least one filter. The system then identifies line Mura candidates by converting the filtered image to a binary image, counting line components along a slope in the binary image, and marking a potential candidate location when a counted number of the line components along the slope exceed a line threshold. Image patches are then generated with the candidate locations at the center of each image patch. The image patches are then classified using a machine learning classifier.