摘要
在图像形态分类与识别系统中，不变性角点检测是很重要的预处理环节。为有效检测角点，并对之精确定位，本文着重介绍一种基于小波变换的不变性角点检测方法．该方法先对轮廓上各点的切向方向进行小波变换，然后提取出变换幅度的局域最大值点作为待定角点．接着，根据待定角点类型建立相对应的模型，并根据模型对待定角点的真伪进行评判．实验表明，上述方法既能正确检测到角点，又能大大提高计算效率．
Invariant corner detection is a most important preprocessing step for image shape classification and recognition. For the purpose of effective corner detection and precise localization, in this paper we propose an efficient invariant corner detection method based on wavelet transform. The method first applies a wavelet transform to the contour direction,then it extracts local magnitude maxima as corner candidates. Afterwards 9 it establishes various models according to different kinds of corner and make corresponding verification based on them. Experiments show that the proposed method not only detects corners accurately but also efficiently.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
1995年第6期112-116,共5页
Journal of Shanghai Jiaotong University
关键词
图像分类
不变性
角点检测
图像识别
image classification and recognition, invariance, corner detection, wavelet transform, image contour
