摘要
针对熊猫光纤端面图像识别,提出了一种亚像素椭圆特征检测方法.通过参数分离和精度逐级递增的模板匹配,将算法的时间复杂度降低为O((log n)2).经实验检验,在低噪声环境下,该算法检测平均误差低于0.04像素;在高噪声和图像不完整的情况下,检测平均误差仍低于0.07像素.
This paper presents a sub-pixel ellipse detection algorithm on image identification of panda fiber endface . Using parameter separation and precision progressive increase method ,the time complexity of the algorithm could be decreased to O((log n)2 ) .The average detection error of this algorithm is less than 0 .04 pixel under low noise image .This number will increase to 0 .07 pixel under the case of high noise and incomplete ellipse .
出处
《微电子学与计算机》
CSCD
北大核心
2014年第7期156-160,166,共6页
Microelectronics & Computer
关键词
亚像素
模板匹配
参数分离
精度递增
sub-pixel
template matching
parameter separation
precision progressive increase
