Wavelet Decomposition for Display Panel Mura Defect Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for detecting Mura defects in display panels, such as TFT-LCD, face challenges in suppressing textured backgrounds without reducing the contrast of the defects, leading to inefficient defect detection.
Innovation Solution
A background suppression method using multi-level wavelet decomposition, coefficient smoothing, and contrast enhancement to separate and maintain the original defect contrast while suppressing the textured background, employing Gaussian low-pass filtering and histogram equalization in an automatic optical detection system.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If Gabor filter method is used to suppress textured background, then textured background is filtered, but contrast of Mura defect is reduced
Solution Approach 1:
The patent applies wavelet decomposition to segment the image into multiple frequency sub-bands (low-frequency sub-band containing background information and high-frequency sub-bands containing defect information). This segmentation allows independent processing of background suppression and defect preservation, resolving the contradiction between filtering textured background and maintaining defect contrast.
Solution Approach 2:
The patent applies different processing strategies to different frequency components: Gaussian filtering is applied to high-frequency sub-bands to suppress textured background, while contrast enhancement is applied to low-frequency sub-bands to preserve defect information. This local quality approach ensures that each sub-band is optimized for its specific function, maintaining defect contrast while suppressing background.
2Object-affected harmful factors
If multiple frequencies and directions of filter convolution are applied, then textured background is filtered in all directions, but defect contrast is reduced
Solution Approach 1:
The patent segments the filtering process by applying Gaussian filtering only to high-frequency sub-bands while preserving low-frequency sub-bands. This segmentation prevents the loss of defect contrast that would result from applying multi-frequency and multi-directional filtering to the entire image, as the low-frequency components containing defect information remain intact.
Solution Approach 2:
The patent applies contrast enhancement specifically to low-frequency sub-bands after filtering high-frequency sub-bands. This local quality processing ensures that defect contrast is restored and enhanced in the regions where it matters most, compensating for any contrast reduction while maintaining effective background suppression.
Data Source
AI summary
The disclosure provides a background suppression method and a detecting device in automatic optical detection for a display panel, wherein the background suppression method includes the following steps: S1: collecting a pure color image of the display panel; S2: performing multi-level wavelet decomposition on the pure color image to obtain a series of high-frequency sub-bands and low-frequency sub-bands; S3: performing coefficient smoothing process on high-frequency sub-bands in multiple directions of each level, and performing contrast enhancement process on each level of low-frequency sub-bands; S4: the processed high-frequency sub-bands and the processed low-frequency sub-bands are subjected to wavelet reconstruction to obtain a defect image after background suppression. The disclosure performs multi-scale and multi-resolution decomposition on the image, and performs texture suppression and image enhancement on the decomposed high-frequency sub-bands and low-frequency sub-bands respectively, and can maintain the contrast of the original defect while suppressing the background texture.