摘要
本文利用小波为工具提出了在多尺度检测图象中屋顶边缘的方法，导出了该处屋顶边缘的最大曲率及其方向的计算公式；并将其结合人脸的特点用以在较复杂背景的图象中确定人脸及其器官（如眼睛）的位置和进行初步的分割，此方法在实验中取得了良好的效果。
In this paper, based on wavelet transform, we propose a new method of multiscale ridge betection in an image, and then the formulas, which are for computing maximal curvature and its direction in there, is derived. Combined with face feature, this method is used for detecting and segmenting faces and its feature (eyesetc.) in complex background. The effectiveness of the new method is confirmed by the experiment in the end ofthis paper.
出处
《信号处理》
CSCD
1998年第4期313-317,共5页
Journal of Signal Processing
关键词
小波变换
多尺度屋顶边缘
目标检测
图象处理
Wavelet transform, Multiscale ridge, Object detection Face detection
