基于小波变换的图像边缘检测12.pdfVIP

2004 7 Jul. 2004 26 7 Systems Engineering and Electronics Vol26 No7 : 00506X( 2004) 王俊卿, 黄莎白, 史泽林, 于海斌 ( 中国科学院沈阳自动化研究所, 辽宁沈阳 00 6) : 提出了一种新的基于小波变换的多尺度多分辨率特 的边缘检测方法分别利用不同尺度小波变换 后的水平方向垂直方向和对角线方向高频信息, 根据李氏指数与小波变换关系, 采用零交叉检测方法和极大值在 不同尺度下传播的特 , 检测出图像在3 个方向的极大值对不同尺度不同方向高频子图像, 首先采用平均值法 确定阈值, 去除震荡噪声, 然后对不同方向高频子图像合成的图像, 根据最大类间方差法, 计算其阈值, 求出极大 值, 最后从这3 个方向的极大值中确定出最大值实验结果表明, 检测出图像的边缘结果非常理想 : 小波变换; 边缘检测; 李氏指数; 多尺度; 多分辨率 :TP39 .4 :A Edge detection image based on wavelet transform WANG Junqing, HUANG Shabai, SHI Zelin, YU Haibin ( Shenyang Instit ute of A utomation, Chinese A c ademy of S ciences , Sh eny ang 110016 , China) Abstract: Based on multiscale and multiresolution of wavelet transform a new algorithm of edge detection has been presented. This algorithm separately utilizes high frequency information of horizontal, vertical and diagonal di rections of wavelet transform under different scales. According to relation of Lipschitz exponent and wavelet transform, the properties of crosszero and propagation under different scales are utilized to determine local max in three direc tions. Firstly, to the subimageof different scales and directions themethod of averagethreshod is utilized to determine the threshold and to denoise. Secondly to the synthesized image of different orientation subimage the method of OSTU is utilized to compute the threshold to max. At last it determines the maximal one of those in three orientation. The experiment presents that the results of edge detection are very good. Key words: wavelet transform; edge detection; Lipschitz exponent; multiscale;