摘要
为实现对胶片弊病的实时识别,提出了一种正交边缘投影快速检测算法。基于对红外胶片图像特点的分析,通过综合运用邻域平均、边缘检测和投影法等图像处理技术,实现了对胶片弊病的快速、准确识别,并据此算法编制了相应的VC++程序。该算法保证实时的关键是将二维图像投影为两个一维信号进行处理,该算法还可根据实际需要实现功能扩展,用于确定弊病的大小及严重程度。通过在乐凯胶片集团公司的在线实验,证明其能够准确检测出各种典型的胶片弊病,处理一帧大小为659×494像素的图像只需20ms。
A fast orthogonal edge projection algorithm is proposed for real-time inspection of film defects on production line. By analyzing the characteristics of infrared images of film, using image-processing techniques including neighborhood averaging, edge detection and projection method, etc., and compiling the VC++ program film defects are quickly identified. In this process, the key is the projection method that transforms two-dimension images into two one-dimension vectors. Furthermore, new functions, such as evaluation of defect area and severity can be added to the algorithm in accordance with the actual requirements. On-line experiments in Lucky Film Corp. proved that the algorithm can detect all kinds of film defects and process a 659+494 image within 20 ms.
出处
《光学精密工程》
EI
CAS
CSCD
2004年第3期323-329,共7页
Optics and Precision Engineering
关键词
图像分割
投影
边缘检测
胶片弊病
image segmentation
projection
edge detection
film defect
