Display Panel Image Quality Inspection with Simulated Moire Removal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Moire patterns occur when images captured by imaging devices inspecting display panels, leading to unreliable image quality inspection due to mismatched pixel arrangements between the display panel and image sensor, affecting the accuracy of inspections.
Innovation Solution
A method using a simulator and deep learning model to generate non-moire and moire images, training a moire removal model to convert moire images into non-moire images, and applying this model to captured evaluation images to remove moire, enhancing inspection reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If an imaging device captures an image displayed on a display panel, then image quality inspection can be performed, but a moire pattern occurs in the captured image due to mismatched pixel arrangements
Solution Approach 1:
The patent creates a virtual copy of the imaging device's pixel arrangement and uses it to generate synthetic moire images through simulation. This virtual model allows the system to learn moire patterns without requiring physical capture, thereby avoiding the information loss that occurs when real images are degraded by moire artifacts.
Solution Approach 2:
The system performs preliminary simulation to generate training data before actual inspection. By pre-generating synthetic moire images that represent various moire conditions, the system prepares the deep learning model in advance, so that during actual inspection, the model can directly remove moire without the degradation affecting real inspection results.
2Reliability
If conventional methods are used to remove moire, then extensive real image capture and processing is required, but this increases time consumption and computational complexity
Solution Approach 1:
The patent replaces the mechanical process of capturing real images with moire patterns and physically processing them with a computational simulation system. The simulation virtually generates moire images by mathematically modeling the interaction between display panel pixels and imaging device pixels, eliminating the need for time-consuming physical image capture and preparation.
Solution Approach 2:
The simulation system varies parameters such as pixel arrangement, viewing angle, and magnification to generate diverse training data. By changing these parameters computationally rather than physically recapturing images under different conditions, the system efficiently creates comprehensive training datasets without increasing time consumption.
3Ease of manufacture
If the pixel arrangement of the display panel and image sensor are fixed, then manufacturing is simplified, but moire patterns occur due to the lattice structure mismatch
Solution Approach 1:
The patent converts the harmful moire pattern into a beneficial training opportunity. By simulating the exact pixel arrangement mismatch that causes moire, the system creates realistic training data that teaches the deep learning model to recognize and remove these specific artifacts, turning the manufacturing simplicity constraint into an advantage for targeted model training.
Data Source
AI summary
A method of inspecting image quality includes receiving a non-moire image generated by a simulator and a moire image generated by the simulator based on the non-moire image, training a moire removal deep learning model for converting the moire image into the non-moire image, from which a moire is removed, using the non-moire image and the moire image, inputting an evaluation image generated by capturing an image displayed on a first display region by an imaging device into the moire removal deep learning model, and removing the moire from the evaluation image through the moire removal deep learning model when the evaluation image includes the moire.