Abstract: Currently，the image defects of LCD products mainly rely on manual detection.This algorithm realizes automatic real-time detection of image defects.Firstly，the real-time images are captured from the buffer of image acquisition card quickly by DirectShow technology.Then the weighted-averaged frame of real-time frames can reduce the bad effect of the bad frames.It removes the noise by Gaussian pyramid sampling.Dynamic threshold value，getting from each RGB 3-channels by recursive iteration，is helpful to detecting the screen rectangle，and then auto-correcting the cell phone position，extracting ROI，completing the image preprocessing.Finally，it detects the contours of defect using the Canny algorithm，and combined with Douglas-
Peucker algorithm and Freeman chain code，it extracts the information of defect，and then tests the image defects of phone screen：the number of bad pixels，geometric distortion，chromatic aberration.This algorithm is real-time and efficient.It can be widely used in LCD products for image defect detection relying on the relevant national standards，and it is worth promoting.

Key words: defect extraction, Region of Interest（ROI） identification, real time detection, phone LCD screen

摘要： 当前液晶屏类产品图像缺陷主要依靠人工检测，该文算法实现了图像缺陷的自动实时检测。首先通过DirectShow技术从图像采集卡缓存区快速获取实时图像；其次，将多帧实时图像加权平均，剔除坏帧，通过高斯金字塔采样去噪及分别从RGB三通道递归迭代获得分割阈值，提取屏幕矩形外框，自动校正手机姿势，继而提取ROI，完成图像的预处理；最后，利用Canny算法检测缺陷轮廓，结合Douglas-Peucker算法与弗里曼链码提取缺陷信息，最终检测手机屏幕图像缺陷：坏点数目，几何失真度，色差。算法实时、高效，依托相关国家标准，可广泛应用于液晶屏类产品的图像缺陷检测，具有一定的推广价值。

关键词: 缺陷提取, 感兴趣区域（ROI）识别, 实时检测, 手机液晶屏