OLED display screen defect detection technology based on skeleton template registration

As a new generation of display devices, OLED (Organic LED) display screens have been widely used in low-power devices such as MP3, mobile phones, and digital cameras as the production process becomes more and more perfect. In the process of automated inspection based on image processing, in order to ensure the quality of products, manufacturers urgently need an effective algorithm to quickly capture and identify various defects in the display screen. Among the various defects of OLED display screens, mura defects (also known as Mura defects) are the most common, most complex, and also the most difficult to detect defects [1-2]. They are mainly characterized by low contrast, blurred boundaries, diverse shapes, and uneven brightness display. Therefore, how to effectively detect mura defects has become a key link in the manufacturing process of OLED display screens.
In recent years, with the development of image processing theory, relevant researchers have proposed many detection algorithms. Yen PingLang et al. proposed a detection method based on background image reconstruction [3], and Kuo CC proposed a method to filter out background images using discrete cosine transform [4]. Due to the low contrast, blurred boundaries and uncertain shapes of the spot defects, coupled with the fact that the brightness of the display screen itself is difficult to achieve completely uniformity and the influence of CCD noise, it is more difficult to extract the spot defects. Conventional threshold segmentation, edge extraction and other methods can no longer effectively extract the spot defects.
To address this problem, this paper proposes a new spot defect detection method. At the system startup stage, an ideal template is created based on the collected image, and the skeleton information of the OLED display screen is extracted using the refinement technology to achieve rapid registration of the template image and the original image, and perform a subtraction operation; then, the threshold determined by the Otsu method (i.e., the maximum inter-class variance method or OTSU algorithm) is used to segment the subtracted image, and the spot defects can be effectively extracted. The algorithm flow is shown in Figure 1.
1 Extraction of the display screen skeleton template
The skeleton (Skeleton), also known as the medial axis (Medial Axis), is an important topological description of the geometric form of the graphic. The skeleton is a linear geometric body. It

In the formula, S(i, j) is the original image, T(i, j) is the template image, and D(i, j) is the image after difference.
In the actual defect algorithm, each point in Figure 4 is used as a control point, and the original image and the small template image are subjected to the difference method to obtain the difference image of the entire image. The difference method detection process is shown in Figure 5. Using this difference detection method, the image after difference processing of the original image shown in Figure 2 is shown in Figure 6.

During the recursive call, t=t+1 until the recursion ends, t=254. The algorithm can improve the computational efficiency by 80% after recursive improvement.
4 Defective image example

Experiments show that the image registration and detection technology based on the display skeleton proposed in this paper can effectively extract the stain defects of the display. In terms of the processing efficiency of the algorithm, using Visual Studio 2008 as the development environment, a 1280×960 resolution image was tested on a notebook with a CPU T6500 and 2 GB memory. The algorithm took 282 ms, of which the skeleton extraction took about 219 ms, the difference method took about 16 ms, and the Otsu algorithm took about 2 ms.
Based on the traditional difference method, this paper improves the search strategy during image registration and proposes a stain defect detection method for OLED display screens based on skeleton template registration. By using the block registration method, the influence of the small angle rotation of the display screen during registration is solved. The stain defects of the display screen can be effectively detected with short time consumption, which can meet the requirements of real-time detection.
References
[1] Zhang Yu, Zhang Jian. Automatic detection system for TFT-LCD stain defects based on polynomial surface fitting [J]. Optoelectronic Engineering, 2006, 33(10): 108-114.
[2] Tang Jian, Wang Dawei. Application of B-spline surface fitting in mura defect acquisition [J]. Modern Display, 2008(89): 24-28.
[3] YEN P L. Automatic optical inspection on TFT-LCD mura defects using background image reconstruction [J]. Key Engineering Materials, 2008, 364/366: 400-403.
[4] KUO C C. Automatic TFT-LCD mura defect inspection using discrete cosine transform-based background filtering and ′just noticeable difference′ quantification strategies [J]. Measurement Science & Technology, 2008, 19(1): 015507-1-015507-10.
[5] Lv Junqi. An effective binary image thinning algorithm [J]. Computer Engineering, 2003, 29(18): 147-148.
[6] Su Xiaohong, He Zhiguang, Ma Peijun. Automatic detection algorithm for micron-level display defects of TFT-LCD[J]. Journal of Harbin Institute of Technology, 2008, 40(11): 1756-1760.
[7] OTSU N. A threshold selection method from gray-level histogram[J]. IEEE Transactions on System, Man, and Cybernetics, 1979, SMC-9(1): 62-66.
[8] Jing Xiaojun, Cai Anni, Sun Jingao. An image segmentation algorithm based on two-dimensional maximum inter-class variance[J]. Journal of Communications, 2001, 22(4): 71-76.
[9] Li Lele, Deng Shanxi. Image block binarization algorithm based on Otsu method[J]. Microcomputer Information, 2005, 21(8-3): 76-77.