摘要
为了取得含噪人耳图像的理想边缘轮廓,以实现人耳识别技术的进一步应用,对小波变换边缘检测方法进行了研究,分析了噪声消除与小波变换尺度之间的关系,详细论述了模局部极大值提取边缘的原理。针对含噪人耳图像的特殊性,阐述了一般去噪和边缘检测方法的不足,并针对这些不足提出了改进方法,首先利用样条小波多尺度分解后,相邻尺度小波系数相乘得到尺度积,然后进一步求得尺度积的模和相角,通过自适应阈值去噪提取图像边缘,取得了较好效果。
To get an ideal edge of a human - ear image with noise for a further application, this paper studies the edge detection methods of wavelet transform, analyses the relationship between noise reduction and scale of wavelet transform, describes the principle of the module maximum edge detection method. According to the particularity of human - ear image with noise, it shows the shortcoming of other edge detection methods and gives an improvement. First the paper makes a multiplication of adjacent scales to get the scale multiplication after wavelet transform, then works out the modules and the phase angles. The image edge is achieved with adaptive thresholds to remove noise components. The result is satisfactory.
出处
《计算机仿真》
CSCD
2008年第1期236-239,共4页
Computer Simulation
基金
重庆市自然科学基金(8596)
关键词
边缘检测
小波变换
多尺度
尺度积
Edge detection
Wavelet transform
Multiscale
Scale multiplication