Abstract: Organic LED（OLED） screen often exists defects which have the characteristics of small size, low contrast, indifferent gray value. These defects will cause detection problem. Traditional thresholding method cannot detect the defects effectively. This paper proposes a multi-image-subtraction method to detect pixel defects of OLED screen. Based on the initial process of median filtering and image enhancement, the image is processed with pixel template extracting and subtraction operation, until a better contrast image is achieved. A K-means clustering method is selected to segment the image and detect the defects from the processed image. Through the tools of Labview and IMAQ Vision software package, this method proposed in the paper is implemented and verified by an OLED image. The test result shows that this method can detect defects effectively and the smallest defects can be 8 μm×8 μm.

Key words: image processing, Organic Light Emitting Display（OLED）, defect detection, subtracting method, K-means clustering

摘要： 有机发光显示器OLED（Organic LED）的内部像素缺陷常由于尺寸小、对比度不高、且灰度与像素轮廓灰度相近等问题，在使用传统阈值分割方法处理时会将像素轮廓保留下来，因而不能达到缺陷的有效检测目的。提出一种多次迭代差影法的OLED屏像素缺陷检测方法。该检测算法在对图像进行中值滤波和图像增强处理后，对图像实施多次像素模板提取和差影运算，获取到对比度较低的缺陷图，运用K-均值聚类方法对图像进行分割，从而较好地实现缺陷的识别。运用Labview和IMAQ Vision软件包工具，编程实现所提出的算法，并通过实际获取的OLED图片验证了方法的有效性。结果表明，该方法能很好地保持缺陷的细节，并能检测到8 μm×8 μm的微小缺陷。

关键词: 图像处理, 有机发光显示器（OLED）, 缺陷检测, 差影法, K-均值聚类